I’ve held most of these opinions myself at one point or another, with roughly the same evidence behind them each time, which is to say a handful of memorable trades and a feeling. So I tested them. Six entry times (9:35, 10:00, 11:00, 1:00, 2:00, 3:00), the same trade at each one: a 15-delta SPX put credit spread, 10 points wide, held all the way to settlement. I priced the entry two ways, at the mid and net of what a marketable order would cost, and I’ll show you both, because for a liquid product like SPX the honest answer is a range. Four years of sessions, and I kept only the days where all six entries were tradeable, so every column is the same set of 820 days.

The short version: the time of day is not where your edge is. It’s where your tail risk is, and those aren’t the same thing.

Subscribe now

The expected return barely moves

_{Grey bar = filled at the mid; teal bar = net of a marketable-order fill. The gap between them is the opening execution cost, an upper bound. Whiskers are the 90% weekly-block bootstrap.}

Left panel is P&L per trade, the number you actually feel. Right panel is that same P&L as a share of what you were risking, which strips out the fact that a later entry is a smaller trade with less premium in it. In both panels the grey bar is a fill at the mid and the teal bar is net of a marketable order — the gap between them is the entry execution cost, and I’ll come back to what that gap really is.

Both are flat. At the mid, every entry is within about a dollar or two of breakeven per trade, call it slightly positive on average. Net of a marketable fill, every entry is slightly negative on average, by about $5 a trade. Either way, on a share-of-risk basis it’s within roughly 1% of zero at every entry, and the confidence intervals all overlap. The 11:00 bar shows a bigger positive number, and if you’d asked me before I ran this I’d have had a story ready about the morning shakeout being done by 11. But its interval crosses zero, it doesn’t repeat as a pattern, and I don’t believe it.

About that fill cost

The “fill” column above is what the trade costs if you send a marketable order — one that crosses the spread to get done right now. It’s built from four years of actual 0DTE prints, measuring how far from the mid trades landed on average. That number is real, but it’s the cost for a specific kind of order, and it’s the upper bound.

If you work a limit at the package mid and wait, you often just get filled at the mid. On the spread widths this trade lives in, roughly half of all real 0DTE prints happen at the mid. My own fills on SPX are usually at or near the mid, and I doubt I’m unusual. The model can’t represent the patient trader who produces those at-mid prints — it can only average them in with the half who crossed. There’s a second reason it runs high, too: it charges each leg of the spread its own slippage, which models legging in, and a real spread order goes to the exchange as one package that’s tighter than the sum of the legs.

So the truth is the band. A trader working limits sits near the mid column, which makes the trade a marginal positive. A trader sending market orders into the close sits near the fill column, a marginal negative. Both are a coin flip.

The win rate creeps up as you go later, from 88% to 90%. It always does that — a later 0DTE trade has less time to get in trouble, so it collects the credit more often. And as with every other study in this series, the win rate is telling you almost nothing. All six entries are a coin flip on expectation once you account for what you’re risking.

If the question was “which entry time makes the most money,” the answer is that it doesn’t matter, and I’d want you to leave with that plainly. But two of those columns move in a way the P&L doesn’t, and that’s the actual content of this piece.

What does move: the tail

_{Share of trades finishing at or near max loss, by entry time. Rank correlation with entry time is −0.94.}

The last column of that table is the share of trades that finished at or near max loss — the spread almost fully in the money, the outcome that actually hurts a premium seller. It falls, cleanly, the later you enter: 7.9% at 9:35, down to 3.8% at 3:00. Roughly cut in half across the trading day.

This one isn’t noise. The rank correlation between entry time and the near-max-loss rate is −0.94, the difference between the first entry and the last clears a straightforward significance test, and it holds in every calendar year in the sample, including the two with real shocks in them. A 9:35 trade and a 3:00 trade have the same expected P&L, but the 9:35 trade gets there through a wider range of outcomes, and the wide end of that range is the end you care about.

Why: there’s less of the day left to move against you

_{Realized vol still ahead of the position at each entry — the square root of the summed squared minute moves from entry to close, median over 1,037 sessions.}

This is the least surprising chart in the study once you see it. It’s the realized volatility still ahead of the position at each entry time — square root of the summed squared minute-by-minute moves from entry to the close, which is the thing a held position is actually exposed to.

At 9:35 there’s about 60 basis points of index movement still to come. By 1:00 it’s 36. By 3:00 it’s 22. The overnight gap is already priced in by 9:35, the first hour is the most active stretch of the session, and the afternoon mostly grinds. Enter later and you’ve handed a chunk of the day’s variance to whoever was holding before you. Fewer big adverse moves left means fewer trades in that near-max-loss bucket. The mechanism is that boring.

The cost of the smaller tail

_{Same 15-delta target at every entry. Later means a smaller credit and a closer strike.}

Nothing’s free. A later entry collects less: the mid credit on the same 15-delta spread drops from about $0.98 at 9:35 to about $0.63 at 3:00, because there’s less time value left to sell. And “15-delta” at 3:00 is a different strike than “15-delta” at 9:35 — with only an hour of life left, a 15-delta strike sits about 16 points from spot instead of 42. You’re selling a closer strike for a smaller credit.

So the afternoon entry doesn’t beat the morning. It just packs less into each trade: less credit, a closer strike, a smaller worst case. If you’re sizing to a fixed dollar risk you’ll write more contracts to get there, which puts you back at a similar total exposure and a similar total expectation, with the individual losses capped lower. Whether that trade-off is worth making is a risk-tolerance question, and I’m not going to pretend the data settles it for you.

The thing I expected to find and didn’t

_{Share of the mid credit given up to execution at entry. Pre-registered prediction that it would fall through the morning: rejected.}

Going in, I assumed a later entry would at least get filled better. Morning quotes are wider, I’ve measured that in other studies, so waiting for them to tighten seemed like it should shave the cost for anyone who does have to cross the spread.

It doesn’t. That marketable-order cost, the share of the mid credit you give up when you take liquidity instead of posting it, sits between 6% and 8% at every entry time, with no improvement through the morning and if anything a slightly worse number in the afternoon. My best guess is that the afternoon spreads stay proportionally about as wide once the option is cheap enough, so the percentage cost doesn’t fall even though the dollar spread does. Whatever the reason, “wait until later for a better fill” isn’t a thing the data supports. I pre-registered that prediction before running the sweep and it was wrong.

This only matters if you’re crossing. If you work a limit at mid, you’re mostly not paying this at any time of day.

The exception: Fed days

_{P&L per trade on the 25 FOMC statement days vs. ordinary days. Small sample — directional.}

Everything above pools every kind of session together. Pull out the 25 days in the sample where the FOMC released a statement at 2:00pm, and the picture changes.

On those days every entry loses money, and it gets worse the later you go in: a 9:35 entry averaged −$50 a trade, a 1:00 entry −$57, a 3:00 entry −$147, with the near-max-loss rate up in the 16–20% range across the board. A position you open at 1:00 or 2:00 is sitting there fully exposed when the statement drops, with no time premium collected yet to cushion it. On ordinary days the same entries run between −$5 and +$9.

The “sell later, smaller tail” logic runs exactly backwards here. On a Fed day, if you’re going to be short into 2:00, you want to have been short since the morning so you’ve banked some decay first — or you want to not have the trade on at all.

Twenty-five days is a small sample and I’d hold this loosely on the exact dollar figures. The direction is not subtle, though, and the mechanism is obvious enough that I’d be surprised if more data softened it.

Where the P&L actually comes from

One 9:35 spread, marked to what it would cost to close through real fills at each point. The 3:00 point is hollow — the wing often has no bid by then, so the cost-to-close overstates.

One more angle on the same position, because it explains the shape of everything above. This tracks a single 9:35 spread and marks it to what it would cost to close, through real fills, at each point in the day.

It’s underwater almost immediately, down about $15 by 10:00, and then it just sits there. Barely moves through the middle of the day. The whole gain shows up in the last stretch into settlement, where it climbs about $22 to finish near breakeven. A 0DTE credit spread held to expiry doesn’t earn steadily. It takes an early mark-to-market hit from the spread and the first hour’s noise, waits, and collects in the final hour.

That’s also why entering later doesn’t help your expectation: you skip the boring underwater middle, but you also skip less of it than you’d think, and the last-hour collection is the same either way.

Does it hold across the ladder?

_{The tail taper at three strike distances. Higher delta means more near-max-loss at every entry, but the same downward slope.}

The tail taper isn’t a 15-delta artifact. Sell the 10-delta strike, the 15, or the 25, and the near-max-loss rate falls through the day in every case. Closer-to-the-money strikes finish near max loss more often at every entry, which is expected, but the downward slope with entry time is the same shape at all three.

What I’d take from this

The clock is not your edge. If you’ve been agonizing over whether to open at 9:35 or wait until 10:30, the data says stop agonizing. On expectation, per unit of risk, it’s the same trade.

A later entry is a lower-variance version of the same trade, not a better one. You collect less, you sell a closer strike, and in exchange your worst outcomes are capped lower. If you’ve been getting hurt by the occasional full-width loss on a morning entry, moving later is a legitimate way to shrink those, as long as you understand you’re shrinking the wins by a similar amount and probably writing more contracts to compensate.

On Fed days, the rule inverts. Don’t open a fresh short into the early afternoon on a statement day. Either be in it from the morning with some decay banked, or sit it out.

Your win rate went up as you entered later, and it meant nothing. Same as always. If the number you watch is the fraction of trades that come back green, you are watching the least informative thing on the screen.

If you’ve moved your 0DTE entry later over the years, has it actually changed your worst months — or just your average one?

What’s next

Every trade in this study was held to settlement. The obvious companion question, whether an exit rule earns its keep on 0DTE once you price what it costs to actually get out, I looked at that, and the answer wasn’t interesting enough to publish on its own. The one I’m building now is 0DTE against 1DTE: whether you get paid enough to carry the position overnight, or whether that first night is where the trade goes wrong.

Subscribe now

Method — and the caveats

Data. SPXW 1-minute consolidated NBBO, June 2022 - August 2026, 1,000+ sessions (June 2022 is when every day expiration on SPX options went live). Delta, implied vol, and the forward are solved from the option chain itself by put-call parity — no model prices anywhere in the trade construction.

One arm: held to settlement. Every position is opened and held to cash settlement, which pays no closing commission and no closing slippage. That’s a real advantage of holding, not a modeling shortcut, and it’s stated rather than smoothed. Only the opening credit is priced two ways, at the mid and net of a fill.

The fill number is a marketable-order cost, and it’s an upper bound. It comes from a model calibrated on four years of actual 0DTE prints: how far from the mid trades landed, on average, as a fraction of the bid/ask spread. That describes an order that crosses the spread. It says nothing about whether a resting limit at the mid would have filled, and on these spread widths, roughly half of all real 0DTE prints happen at the mid, so plenty of them would. The model also charges each leg its own slippage, which models legging into the spread; a real spread order trades as one package that’s tighter than the sum of the legs. Both of those push the fill number high. A patient limit-order trader is closer to the mid column; a market-order trader is closer to the fill column. The commission is $0.65 per contract per leg, carried in both columns.

The balanced panel. A session counts only if all six entry times produced a tradeable fill-priced spread that day — 820 of 831 candidate sessions. Re-running on every available session instead moves nothing of substance.

The duration question. A 9:35 trade is a six-and-a-half-hour position; a 3:00 trade is one hour. I did not add an artificial fixed-holding-window arm to force them equal, because a held-to-settlement trade is what people actually do and the entries are nested anyway — the gap between two consecutive hold-to-settlement entries is the contribution of the hour between them. The per-hour-of-risk figure is in the results file for anyone who wants it; I led with per-trade and per-max-risk because dividing a near-zero P&L by a one-hour holding period mostly amplifies noise.

FOMC days. Tagged from the Federal Reserve’s published meeting calendar, 25 statement days in the fill-priced window. CPI, PPI, and payrolls all release at 8:30am, before every entry in this study, so they move the whole session but don’t create an entry-timing effect the way a 2:00pm statement does. They’re not split out here.

Four years, not decades. The 0DTE series buys precision with real minute quotes and pays for it with a short history — 2022 through 2026, bull-heavy, two shocks. The tail taper holds in every year of that sample. It has not been tested against a market that trends down for a year, because this sample doesn’t contain one.

THETA DESK

Subscribe now

TL;DR

Every gauge of how expensive options are fell this week. IV rank — where each index’s implied vol sits in its own past year — dropped on all three (S&P 9.4 → 5.5, Russell 8.4 → 3.2, Nasdaq 29.3 → 16.6). The VIX fell from 15.13 to 14.43. Every weekly priced range narrowed (S&P ±1.22% → ±1.05%, Russell ±1.61% → ±1.54%, Nasdaq ±1.92% → ±1.60%).

And every variance risk premium got wider — the gap between what options charge and what the index has actually been delivering. The S&P’s went from 1.15× to 1.36×, the Russell’s held near 1.22×, and the Nasdaq’s turned positive for the first time since early August (0.98× → 1.10×).

That reads like a contradiction and isn’t. Sellers aren’t being paid more this week. They’re being paid less. The cushion widened because 20-day realized volatility fell faster than implied did, as the violent first week of August, the week all three indexes broke every range at once, rolled out of the trailing 20-session window. It’s a backward-looking improvement, and it keeps improving mechanically into next week no matter what price does.

The calendar: the August jobs report lands Friday, September 4, at 8:30 AM ET — the same session this week’s ranges settle, and the last major labor-market read before a September 16 Fed meeting the market now prices as a near-coin-flip on a rate hike (roughly 55–59% odds after Chair Warsh’s Jackson Hole speech). ISM manufacturing Tuesday, JOLTS Wednesday, ADP and ISM services Thursday.

The running record of where each index closed versus its Sunday range stands at 18 of 27 (66.7%) after last week’s third clean sweep.

The regime right now

Last week closed with a third straight 3-for-3 scorecard — Nvidia and the new Fed Chair’s Jackson Hole keynote both landed inside the week and every index still finished inside its range. This week opens with vol coming off across the whole curve.

• VIX 14.43 — down 0.70 from last Sunday’s 15.13. Lower third of normal.

• Term structure: 11.22 → 14.43 → 17.48 (9-day → 30-day → 3-month). Still contango. The 9-day-to-3-month gap widened slightly to 6.26 points from 5.92 last Sunday — but that’s the front end falling (VIX9D down 1.36) faster than the back (VIX3M down 1.02), not the back rising. Last Sunday the front had spiked on the run-up to Jackson Hole; this week it gave most of that back.

_{The curve, last Sunday vs. now. The whole thing shifted down; the front gave back its pre-Jackson-Hole spike.}

The receipt from last week: the S&P closed net +0.49% on 3.14% of travel (15% efficiency), the Russell net −1.51% on 4.03% (37%), the Nasdaq net +0.43% on 5.92% (7%). The Nasdaq covered more than three times its whole weekly range in true distance to finish essentially flat — 7% efficiency, among the lowest of any settled week this summer. Only the Russell actually went anywhere, and it went down.

_{Last 21 sessions per index, cumulative % change, against realized-vol bands built from each index's own HV20 — backward-looking, not a forecast. Last 5 sessions shaded.}

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

_{Variance risk premium, prior Sunday vs. now, ranked against each index's own past year. Every cushion widened — but implied vol fell on all three, so the widening is realized vol dropping faster, not options getting bid up.}

A few things worth reading off this slowly.

Every headline number fell and every cushion widened — opposite directions. That combination only happens when realized volatility is falling faster than implied, which is exactly what a trailing 20-day window does after a violent stretch ages out of it. The first week of August, the 0-for-3 week with back-to-back +2.5σ sessions, dropped out of all three windows this week. The cushion didn’t get better because the market got nervous; it got better because the arithmetic rolled forward.

The Nasdaq’s flip is the headline, and it deserves the un-flattering version. A run of readings below 1× that goes back to early August has ended. But implied vol on the Nasdaq fell this week, from 21.98 to 19.92. It turned positive because HV20 fell more — 22.31 to 18.12. Friday’s Debrief flagged this flip at the close; it holds at Sunday’s read. The negative streak didn’t end because the market started paying up for Nasdaq risk. It ended because the calendar turned a page.

That distinction matters for the week ahead. HV20 will keep declining as more of that early-August stretch rolls off, whatever price does over the next five sessions. A cushion that widens because its denominator is shrinking on a schedule is not the same as one that widens because the market is nervous — and only the second one tells you anything about the week to come.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike the VIX they price this specific week, catalysts and all.

_{This week's ±1σ cones through Friday 9/4. Every cone is narrower than last week's — undoing the prior week's widening in one move.}

Every cone is narrower this week, on all three lanes — undoing last week’s widening in one move. That tracks the cheaper implied vol in the lane table above; narrower cones are the direct consequence of paying less for the same length of time.

But note what the straddle is and isn’t pricing. Friday’s jobs report sits inside this window, and this week’s cone is roughly 14–17% tighter than last week’s — a week whose big events were Jackson Hole, Nvidia and a PCE/GDP cluster. Either the market reads an August payrolls print as lower-variance than that, or the event premium for it hasn’t been added to the front of the week yet. Wednesday’s Recalibration is the check on which.

The same range, against where each index has actually been trading:

A month of daily candles per index, then this week's priced range fanned out from Friday's close — same ±1σ weekly expected move as the table above, not a new number.

What it means for premium sellers

Educational framing, not a trade — conditions, not signals.

• A cushion that widens on a shrinking denominator is not a forward signal. Every variance risk premium on the board looks better than last Sunday, and the entire improvement is 20-day realized vol falling as August’s violent open ages out of the window. That process continues next week on its own; it says nothing about whether this week delivers.

• The Nasdaq’s month-long stretch of negative readings ending is real, but read the cause. Implied vol on the Nasdaq fell this week. The flip to positive is a denominator story — the last month’s delivered vol dropping — not a “the market is finally paying for tech risk” story.

• A data print on the settlement session is a different risk than one mid-week. August payrolls lands Friday at 8:30 AM, hours before these ranges settle against the close. Anyone holding through Friday is holding through the number with no session left to adjust.

• This jobs report carries more than the usual weight. It’s the last major labor read before a September 16 Fed decision that, after Chair Warsh’s Jackson Hole remarks, the market prices as a near-coin-flip on a 25-basis-point hike — an unusual direction for that debate to run. Confirm the odds and the calendar against your own sources.

• I don’t know whether the cheaper cones mean the market genuinely sees a calm week or just hasn’t priced Friday yet. The honest read is that this week’s straddle is about 15% tighter than last week’s while still containing a top-tier event. That gap resolves one way or the other by Wednesday.

The risk lens

What a careful seller actually watches this week:

• The August jobs report, Friday, Sept 4, 8:30 AM ET. On this week’s settlement session, and the last major labor-market read before the September 16 Fed meeting.

• The September Fed decision it feeds. Priced as a near-coin-flip on a 25-basis-point hike (~55–59%) after Chair Warsh’s Jackson Hole remarks — up from roughly a third before. The Fed’s communication blackout begins the day after this week ends, so this week’s Fedspeak is the last before the decision.

• A full slate of labor and activity data ahead of Friday — ISM Manufacturing (Tue), JOLTS job openings (Wed), ADP and ISM Services (Thu). Each is a partial read on the same question Friday answers in full.

• Every variance risk premium on the board, wider this Sunday but on a mechanically shrinking realized-vol denominator. Worth a fresh read next Sunday for whether implied or realized does the moving from here.

• The running cone record, 18 of 27 (66.7%). Every miss in this record so far has been a whole-week miss — three weeks carry all nine failures, six weeks are clean, nothing in between. Worth watching whether a jobs-report-on-settlement-day week is calm or breaks all three at once.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them. Markets are open all five days; Labor Day is the following Monday.

• Mon 8/31 — Chicago PMI, 9:45 AM ET. Dallas Fed Manufacturing, 10:30 AM ET. Month-end.

• Tue 9/1 — ISM Manufacturing PMI, 10:00 AM ET. S&P Global final Manufacturing PMI. Construction Spending, 10:00 AM ET.

• Wed 9/2 — JOLTS Job Openings, 10:00 AM ET. Factory Orders, 10:00 AM ET. Fed Beige Book, 2:00 PM ET.

• Thu 9/3 — ADP Employment Change, 8:15 AM ET. Weekly jobless claims, 8:30 AM ET. ISM Services PMI, 10:00 AM ET. Trade balance.

• Fri 9/4 — August jobs report — nonfarm payrolls and the unemployment rate, 8:30 AM ET. This week’s ranges settle against Friday’s close.

_{The same week, laid out day by day. A slate of labor data building to Friday's payrolls print, which lands on the cone's own settlement session.}

A week of labor data building to Friday’s payrolls print, which lands on the settlement session and feeds a Fed meeting that’s now a live call. I don’t know whether the market prices that in gradually or holds it to Friday. That’s what Wednesday’s Recalibration is for.

Your turn: Every variance risk premium in this issue widened this week, and every bit of that came from realized volatility falling as early August rolled out of the 20-day window — not from options getting bid up. Does a cushion that improved for mechanical reasons change how you’d size into a jobs report that settles on Friday’s close, or do you treat “wider is wider” regardless of why? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

— Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday’s Setup was built on one structural fact: both of the quarter’s biggest scheduled catalysts landed inside this week’s pricing window, and one of them landed on the settlement session itself. Nvidia reported Wednesday night. New Fed Chair Kevin Warsh gave his first Jackson Hole keynote as chair on Friday morning, hours before the week’s ranges settled against Friday’s close.

Both landed, and neither was quiet. Nvidia’s stock rose about 8.7% Thursday, its biggest single day since April 2025, on revenue of roughly $96.2B (up about 106% year over year). Warsh said Friday that recent inflation readings haven’t shown meaningful improvement, and bets on a September rate hike spiked to roughly 57% by one widely reported read of futures pricing.

And all three indexes still closed inside the range their options priced a week earlier. The S&P at 7,711.76 (+0.40σ), the Russell at 2,972.37 (−0.93σ), the Nasdaq at 29,433.43 (+0.22σ). That’s the third clean sweep in this record, and the running total climbs to 18 of 27 — 66.7%, up from 62.5% last Friday and now within a point of the roughly 68% a ±1σ range is designed to deliver.

The margins are where the week actually lived. The Russell cleared the floor of its range by 3.2 points, 0.11%. The S&P traded 3.7 points above its ceiling intraday Friday morning and handed the whole move back before the close. And the Nasdaq covered 5.92% of ground, 3.1× its entire priced week, to finish 0.43% higher. A scorecard that reads three-for-three describes an endpoint. It does not describe this week.

The week in vol

Sunday’s issue flagged that the week was never discounted: the straddle expiring Friday charged 78–80% of the original five-session range against the 77.5% that flat-vol scaling implies, meaning the market priced essentially zero event premium for Nvidia and the keynote. That call gets its answer here.

Implied vol rose into Monday and then fell through both events, not into them.

• VIX9D: 12.58 → 11.22. Up to 14.07 by Monday’s close, the largest single-session move of the week, then down every session after, finishing 2.85 points below where it peaked. Net −1.36.

• VIX: 15.13 → 14.43. Net −0.70, the same shape at a quarter the amplitude.

• VIX3M: 18.50 → 17.48. Net −1.02. This is the notable one: the back end has barely moved in every week this dataset has tracked, and this week it fell more than the 30-day gauge did.

_{The front end bid up into the event risk on Monday, then deflated through both events. The 9-day-to-3-month gap: 5.92 Sunday, 4.49 Monday, 6.26 Friday.}

The 9-day-to-3-month gap: 5.92 → 4.49 (Monday) → 6.26 (Friday). It compressed hard as the front end bid up into the event risk, then re-steepened past where it started once the front end collapsed. At 6.26 it’s the fourth-widest reading logged, short of the 7.85 record set two Fridays ago, and the mechanism this time is the front end deflating rather than the back end firming.

All three IV ranks fell back toward, but not through, the record lows set two weeks ago: S&P 9.4 → 5.5 (its record is 4.4), Russell 8.4 → 3.2 (record 0.5), Nasdaq 29.3 → 16.6. Two of the biggest scheduled events of the quarter cleared, and the market’s answer was to charge less than it had before them, on all three names.

The scorecard: did the market stay inside the range?

The range settles on Friday’s close. That’s the only bell that counts for this table.

_{Each bar is one session's actual high-to-low range against Sunday's ±1σ range. The S&P's Friday high pokes above the ceiling; the Russell closes on its own low, 3.2 points above the floor.}

Two of those three finished close enough to the edge to be worth stating in points rather than percentages, and both of them got there on the same session.

The Russell closed at 2,972.37 against a floor of 2,969.2 — 3.2 points of room, 0.11%. It also closed at its own low for the day: Friday’s session low and Friday’s close are the same number. There was no bounce off the bottom, just a close that happened to land above the line. At −0.93σ this is the second-closest Inside grade in the record; the July 5th week’s S&P at +0.99σ is the only one that came closer.

The S&P went the other way and briefly failed. Friday’s intraday high of 7,771.48 sits 3.68 points above the 7,767.8 ceiling, about +1.04σ, outside the range, reached in the post-keynote rally before stocks turned. It closed 59.72 points lower at +0.40σ. On a close-only scorecard that never happened. It happened.

Neither of those is a scoring error; the range is defined on the close and every index respected it there. But two 3-point margins in opposite directions, on the same session, on the same catalyst, is a fair description of how much slack this “clean sweep” actually had.

What actually did it

Monday was risk-off ahead of the week’s events, with all three closing lower and the front-end vol gauge putting in its high for the week. Tuesday and Wednesday were flat holding sessions: the S&P’s two closes finished 1.6 points apart, and Wednesday’s macro cluster (durable goods, PCE, and the second estimate of Q2 GDP) passed without a directional break. Wednesday night Nvidia reported: revenue about $96.2B, up roughly 106% year over year, EPS $2.22. Thursday was the payoff. Nvidia’s stock rose about 8.7%, its biggest day since April 2025, chip names rallied broadly, and the Nasdaq-100 gained 1.43% while the S&P added 0.72%. Friday brought Warsh’s keynote: an initial rally that took the S&P through its ceiling, then a full reversal as the hike-odds repricing sank in. The S&P finished −0.25% on the session, the Nasdaq −0.70%, and the Russell −1.39%, the worst of the three, which is the ordinary shape when rate-hike odds jump. Those are the drivers as reported at the time; confirm against your own sources before leaning on any specific figure.

_{Friday alone delivered 75%, 92% and 65% of what each index's options had priced for the entire five-session week.}

The final session alone covered most of what the entire week had been priced to move. Friday’s high-to-low range was 0.91% on the S&P, 1.48% on the Russell, 1.24% on the Nasdaq, against whole-week priced ranges of ±1.22%, ±1.61% and ±1.92%. That’s 75%, 92% and 65% of a five-day expected move delivered in one session. The week’s ranges held, and they held on a day that used up nearly all of the room they had left.

The running record, nine Sundays priced and nine Fridays graded:

_{Nine Sundays priced, graded. 18 of 27 inside — 66.7%, and every miss so far has taken all three indexes at once.}

18 of 27 — 66.7%, against a ±1σ design that expects roughly 68%. Last Friday this read 62.5%. Twenty-seven observations is still small, but the shape of this record is now clearer than the headline percentage. Every single miss in it has been a whole-week, all-three-indexes miss. Three weeks accounting for all nine failures, six clean weeks, and nothing in between. That is a very different risk profile from nine independent per-index misses scattered across nine weeks, and it's the part of this table I'd actually plan around.

What the seller actually collected

Same-method comparison as always: what each lane charged Sunday, against what it’s actually delivered on the standard 20-day realized window.

_{Every lane was paid more than the 20-day window delivered — including the Nasdaq, whose ratio had been below 1.00× for over a month}

The Nasdaq’s variance risk premium is positive for the first time in over a month. It had printed negative on five straight reads: 2nd percentile three Sundays ago, 3rd, then 11th, then 17th at Sunday’s baseline. It’s now 1.10×, 32nd percentile. The implied side of that name has been getting outrun by realized since late July, and this week it stopped.

That flip deserves an honest accounting of why, because it is not the flattering version. Implied on the Nasdaq fell this week, 21.98 → 19.92. The ratio improved because realized fell faster: HV20 dropped 22.31 → 18.12. That is largely the early-August violence, the record-setting Monday and Tuesday of the August 2nd week, rolling out the back of a trailing twenty-session window. HV20 is not a measurement of these five days; it’s a measurement of the last twenty, and this week’s arithmetic includes the disappearance of sessions that were never part of it. The same effect runs through the S&P, whose HV20 fell 13.17 → 10.58 and whose ratio is the fattest on the board at 1.43×.

So the fair statement is narrow: on the standard measure, every lane was paid more than it delivered this week, and the Nasdaq’s month-long deficit closed. Whether it stays closed depends on whether the next twenty sessions look like these five or like the first week of August, and the window rolling forward will answer that with or without any new price action.

The path: 5.9% of ground to gain 0.4%

_{Total distance covered against net change, the closed week. Two of three indexes finished within half a percent of where they started.}

Efficiency is net move divided by total distance traveled — how much of the walking turned into progress.

The Nasdaq’s 7% is the lowest efficiency of any settled week in this record except one (July 31st’s 4%). It covered 5.92% in true range, 3.1× the entire ±1.92% its options priced for the week, and finished 0.43% higher than it started. Monday down, Tuesday and Wednesday flat, Thursday up 1.43% on Nvidia, Friday down 0.70% on the keynote: a week that went somewhere twice and came back both times.

This is the split the endpoint scorecard cannot show you. The range prices where you finish; travel prices what you had to sit through. Every index covered 2.5–3.1× its priced week in distance while two of the three ended within half a percent of where they began. A position sized to the range was fine on Friday afternoon. A position that got adjusted on Monday’s risk-off, or Thursday’s Nvidia gap, or Friday’s post-keynote reversal, paid for each of those moves in a week whose endpoint says nothing happened.

The Russell is the exception, and it’s the useful one: 37% efficiency, the highest of the three, because it actually went somewhere — down 1.51%, most of it in the final session. Its cushion against the floor came from the size of the range it was given, not from the calm of the path.

What it means for premium sellers

Education, not signals — conditions, not calls.

• A three-for-three week and a comfortable week are not the same claim. The Russell cleared its floor by 3.2 points and the S&P spent part of Friday morning trading above its ceiling. The scorecard grades the close, correctly, but the distance between “graded Inside” and “was never in question” was about a tenth of a percent on one of these three names.

• Sunday’s issue said the week carried no event premium, and the settlement agreed. The Friday straddle charged essentially flat-vol scaling for a week containing Nvidia and a Fed Chair’s debut keynote, and the ranges held anyway. That’s one observation, not a rule about event pricing, but it’s the honest scoring of a specific call this newsletter made in advance, and the call was right.

• The Nasdaq’s negative variance premium closed, mostly through the calendar. Five straight negative reads ended at 1.10×, 32nd percentile. The driver was a trailing 20-day window shedding the first week of August, not implied re-pricing higher — implied actually fell. Treat a ratio built on a rolling window as partly a statement about which sessions are in the window.

• Every miss in this record has been a whole-week miss. Nine graded weeks, three of them failures, and all three failed on all three indexes at once. Six weeks graded three-for-three. If your sizing assumes per-index breaks arrive independently, this record says they haven’t yet: the failure mode has been correlated every single time, which is exactly the scenario a multi-index premium book is least diversified against.

• I don’t know whether Friday’s reversal is the start of a repricing or a one-session digestion of a hawkish speech. September rate-hike odds moved a long way on one keynote, the next FOMC decision is more than two weeks out, and implied vol fell on every gauge into that repricing. Those two things point in opposite directions and I can’t reconcile them from this week’s data. That’s the open question going into Sunday.

The week ahead

The full read is Sunday, so this is the shape only, and I’d rather price it off Sunday’s data than guess here.

Next week is a jobs week. By the sources checked, ISM Manufacturing lands Tuesday, ADP employment Wednesday, ISM Services Thursday, and the August employment report Friday morning — the same settlement-session placement the keynote had this week, and the first labor print since Warsh told Jackson Hole the inflation fight isn’t over. The September FOMC decision (Sept 15–16, with the quarterly projections) sits well outside next week’s range, but it is the thing next week’s data will be read against. Confirm every date and time against your own economic calendar.

Implied vol goes into it cheaper than it came out of last week — VIX 15.13 → 14.43, IV ranks of 5.5 / 3.2 / 16.6 against last Sunday’s 9.4 / 8.4 / 29.3 — while realized, on the trailing window, has fallen faster still. Whether that combination survives a payrolls print is Sunday’s question.

_{Next week's calendar. The August employment report lands Friday morning, on the settlement session again. Confirm dates and times against your own calendar.}

Your turn: Nine weeks into this record, every failure has been a whole-week failure — three weeks where all three indexes broke together, six weeks where none of them did, and nothing in between. The headline hit rate (18 of 27, 66.7%) is almost exactly what a ±1σ range is designed to deliver, but the shape underneath it isn’t random-looking at all. Does the correlated failure pattern change how you’d size across multiple indexes, or is nine weeks still too few to read anything into? Reply — I read and answer every one, and the best reads make it into Sunday’s Setup.

Theta Desk

Subscribe now

Subscribe now

TL;DR

Look only at where the indexes closed Tuesday and this week has not happened yet. The S&P 500 sits 0.04% above Friday’s close. The Russell is down 0.26%, the Nasdaq down 0.34%. In the units Sunday’s ranges were drawn in, that’s +0.03σ, −0.16σ and −0.18σ — all three comfortably inside, the flattest midweek reading this series has logged.

Look at the distance instead and it’s a different week. The Nasdaq covered 2.55% of ground getting to that 0.34% — a third more than its entire week was priced to move. Down 0.97% Monday on a semiconductor selloff, back up 0.64% Tuesday. Only 13% of the walking turned into progress.

And the market has not discounted what’s left. The options expiring Friday still charge 0.98% / 1.26% / 1.49% for the three sessions remaining — 78–80% of the whole five-session range, when flat-vol arithmetic alone would give 77.5%. Nvidia reports tonight, the PCE cluster prints this morning, and the new Fed Chair gives his first Jackson Hole keynote Friday, and none of that is priced above the ordinary passage of time.

Where we are vs. Sunday’s ranges

Sunday’s expected-move ranges run through Friday’s close — the market’s own one-standard-deviation estimate, roughly a 2-in-3 chance of finishing inside. Two of five sessions are gone. In the chart, the band is Sunday’s range, the solid bars are each session’s actual high-to-low, the markers are the closes, and the shaded block on the right is what the market is currently charging for the three sessions still to come.

_{Sunday's ±1σ ranges through Friday 8/28, with each session's actual high-to-low drawn as a bar. All three closes are inside, and no session bar touched an edge. The shaded block is the Wed–Fri move quoted from the 8/28 straddle — on all three, breaking out from here costs more than the entire remaining budget.}

This is the mirror image of last week, when all three indexes had already closed below their floors by Tuesday. This week all three are holding, and holding with room. Breaking out of Sunday’s range by Friday now takes more than everything the market still has priced — 1.20× on the S&P, 1.08× on the Russell, 1.07× on the Nasdaq. Last Wednesday, getting back inside cost 0.30–0.44× of the remaining budget; this week getting out costs more than the whole thing.

Friday 8/28 is a listed expiry on all three indexes, so that denominator is quoted from the straddle covering exactly the three sessions that are left, not approximated by scaling the weekly number down. That matters in a week like this one: scaling assumes vol is flat across the term, which in an event week it usually isn’t.

The closest anything came to an edge was Monday, intraday: the Nasdaq’s low of 28,875.96 sat 0.45% above its floor before Tuesday took it back. The S&P and Russell never got within 0.7% of theirs. None of those are closes, and these ranges settle on Friday’s close, not on the lows along the way.

The running record on this series stands at 15 of 24 (62.5%) across eight Sundays, against a ±1σ design that expects roughly 68%. Two of those eight Sundays broke all three names at once.

What actually did it

Monday was a chip story. A semiconductor selloff (Micron down about 5.8%, AMD and Broadcom off more than 3% and 2%) dragged the Nasdaq down 0.97% while the Dow finished higher, with new US sanctions on Iran and stalled US–Canada trade talks in the background. Tuesday the Conference Board’s consumer confidence index fell to 89.4, its lowest since January, led by a 7.8% drop in the expectations component — and all three indexes rose anyway, into Nvidia’s print. Those are the drivers reported at the time; confirm the specific figures against your own sources.

Down on a sector shock, up on a soft macro print. The two sessions very nearly cancelled.

The distance, and how straight it was

Efficiency is net move divided by total distance covered — how much of the walking turned into progress. At 4–18% this is about as close to pure chop as this series has measured. Last Wednesday’s two sessions ran 70–92%; the same two-day window, one week apart, produced the opposite reading on all three names.

Set the travel against each index’s own five-session range and the Nasdaq separates from the other two: 2.55% traveled against a ±1.92% range is 1.33× the whole week’s priced move, spent in two of five sessions. The Russell used 0.89× of its range, the S&P 0.76×. Distance is not direction. The Nasdaq spent the most of it and finished the two sessions essentially where it began.

Two sessions is a thin window and the decimals shouldn’t be over-read: the S&P’s 4% efficiency is arithmetic on a net move of three index points, and it would swing wildly on a single ordinary session. The comparison that holds up is the ordering. The Nasdaq covered nearly three times the S&P’s ground for a net move in the same direction and of the same rough size.

What’s changed in vol

Same-method comparison — each gauge against Sunday’s read, nothing rebased.

_{The curve stayed in contango and flattened from both ends at once: the 9-day gauge up 0.87 and the 3-month down 0.29, closing the gap from 5.92 to 4.76 points.}

The curve is still in contango and still flattening, but this time from both ends at once — the front end up 0.87, the back end down 0.29. Most of the flattening readings this series has logged came almost entirely from one side: last week’s 2.17-point compression was a front-end shock, and the four straight flattening weeks in late July ran through the front too. A simultaneous squeeze from both directions is a less common shape, and I’d rather flag it than explain it. Two sessions is not enough to separate a real term-structure shift from noise in two different gauges.

IV ranks barely moved. The S&P at 11.3 and the Russell at 9.4 are still bottom-decile against their own past year — Sunday’s near-record-cheap starting point has not been repriced by two sessions of chop. The Nasdaq’s actually ticked down, the only gauge in this table pointing at less risk ahead than Sunday saw, on the day its largest constituent reports.

The cushion

The variance risk premium is what implied vol charges measured against what the index has actually delivered — the cushion behind a premium sale, as opposed to the headline size of the credit.

The three-Sunday thickening streak has stalled. Sunday’s Setup made a point of every cushion improving for a third consecutive week; two sessions later all three are unchanged to within three percentile points, one up and two down. Call that a flat read rather than a reversal. The pattern that issue flagged did not extend through the first half of this week.

The Nasdaq’s is still negative, a fifth straight check across three weeks. Its options are charging 21.79 against 22.26 actually delivered over the last twenty sessions, and that has now survived a range break, a full recovery week, and a chip selloff without flipping positive.

One thing that table hides, said plainly. The 10-day realized numbers are far below the 20-day ones on all three names — S&P 8.19 against 13.27, Russell 13.04 against 15.97, Nasdaq 13.70 against 22.26. Measured on the last ten sessions instead of the last twenty, the Nasdaq’s premium isn’t negative at all: it’s 1.59×, and the S&P’s is 1.89×. The negative Nasdaq reading is being carried by sessions that are about to roll out of the window. I’m not swapping the metric mid-issue. HV20 is the number this series has used every week, and changing windows because one of them tells a nicer story is how a track record stops meaning anything. But the choice of window is doing real work in that column this week.

The adjustment

• Being inside the range is not the same as the range being quiet. All three indexes are comfortably inside Sunday’s bands, and the Nasdaq got there having covered a third more distance than the whole week was priced for. An endpoint measure and a path measure disagree completely this week, and a book that has to be managed daily lives on the second one.

• The remaining move is priced at flat-vol scaling, catalysts and all. Three sessions left, 78–80% of the original five-session range still charged, against the 77.5% that √-of-time gives with no event premium whatsoever. That’s the observation; whether it’s the market judging these events ordinary or the market not paying attention is not something this number distinguishes.

• A wide cushion to the edge is a fact about Tuesday, not about Friday. Breaking out now costs more than everything still priced — but all three of the week’s highest-reach events sit in front of that pricing, not behind it.

• The Nasdaq’s negative variance premium is a five-read pattern, and the window is doing some of the work. On 20 sessions it’s 0.98×; on 10 it’s 1.59×. Both are true, they describe different populations, and only one of them is the metric this series commits to.

• I don’t know whether the two-session chop is the week going quiet or the week waiting. A market that trades sideways ahead of a stacked calendar and one that has genuinely stopped caring look identical until the catalysts land. Friday’s Debrief settles it; nothing between here and Thursday will.

What’s left on the week’s calendar

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Wed 8/26 (today) — PCE, Durable Goods, and Q2 GDP (2nd estimate), all 8:30 AM ET. Consensus has GDP revised to roughly 1.5% annualized from 2.1%, and core PCE steady near 3.3% year over year. Nvidia reports after the close, around 4:20 PM ET.

• Thu 8/27 — Weekly jobless claims, 8:30 AM ET. The Jackson Hole symposium opens (Aug 27–29), themed on financial innovation and payments.

• Fri 8/28 — New Fed Chair Kevin Warsh’s first Jackson Hole keynote as chair. This week’s ranges settle against Friday’s close, hours later.

What's left: three consecutive sessions each carrying a high-reach event, with the last one landing the same morning this week's pricing expires.

Three separate high-reach events in three consecutive sessions, and the last one lands the same morning the week’s pricing expires. Sunday’s issue said it didn’t know whether that keynote would get priced in gradually or held to the last session. Two sessions of data have not answered it: the front end moved 0.87 of a point, which is movement, but nothing that looks like a week repricing itself around a Friday event.

Your turn: The options expiring Friday charge 78–80% of the whole week’s original range for the three sessions that are left — essentially flat-vol scaling, with the PCE cluster, Nvidia, and the Fed Chair’s first keynote all still ahead. Does that read to you as the market genuinely not expecting much from those three, or as time-scaling being a bad model for a back-loaded week? Reply — I read and answer every one, and the best reads make Friday’s Debrief.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

Two Sundays ago, Jackson Hole and the new Fed Chair’s first address as chair sat one week past where the cones settled — a real catalyst, but not that week’s risk. That’s no longer true. New Fed Chair Kevin Warsh delivers his first Jackson Hole keynote Friday morning, August 28 — the same session this week’s priced ranges settle against Friday’s close. Nvidia reports Wednesday night, the week’s other concrete single-name catalyst.

The vol backdrop: all three IV ranks came off last week’s near-record lows (S&P 4.4 → 9.4, Russell 0.5 → 8.4, Nasdaq 21.5 → 29.3), and every variance risk premium thickened for a third straight Sunday — the same pattern Wednesday’s Recalibration flagged mid-week and Friday’s Debrief confirmed held through the close. The Nasdaq’s is still the one lane where it’s negative: 0.98×, 17th percentile, unchanged since Friday’s close but a fourth straight negative read across three weeks.

Every cone is wider than last week’s on all three names — the first time that’s happened in several editions, after a run of narrowing weeks. The running cone record stands at 15 of 24 (62.5%) after last week’s second 0-for-3.

The regime right now

Last week’s Debrief closed out a rough one — the second 0-for-3 in this record, and every cushion thickened anyway. This week opens with vol still elevated from that move, but the front end has already started giving it back.

• VIX 15.13 — up 0.88 from last Sunday’s 14.25. Middle of normal.

• Term structure: 12.58 → 15.13 → 18.50 (9-day → 30-day → 3-month). Still contango. The 9-day-to-3-month gap narrowed to 5.92 points, down from 7.85 last Sunday — the widest we’d logged in eight weeks of tracking. The front did essentially all of the moving, the opposite direction from every week before it: VIX9D up 1.97, VIX3M up just 0.04.

_{The curve, last Sunday vs. now. The gap that took seven weeks to build unwound in five sessions.}

The receipt from last week: the S&P closed net −1.43% on 3.59% of travel (40% efficiency), the Russell net −1.65% on 5.23% (32%), the Nasdaq net −2.45% on 5.75% (43%). More directional than the chop this series has flagged as sub-30%, but with real give-back on all three — consistent with a week that opened one way, rebounded midweek, then sold off again into Thursday's close.

_{Last 21 sessions per index, cumulative % change, against realized-vol bands built from each index's own HV20 — backward-looking, not a forecast. Last 5 sessions shaded.}

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

_{Variance risk premium, prior Sunday vs. now, ranked against each index's own past year. Every cushion widened again — the third straight Sunday.}

A few things worth reading off this slowly.

Every headline number rose off last week’s near-record lows, and every cushion moved the same direction — thicker. That’s the ordinary pattern, unlike two Sundays ago when the S&P’s headline and cushion split apart. This time IV rank and VRP percentile agree on all three names: cheaper two weeks ago, richer now, better cushioned now.

The Russell now has the fattest cushion on the board — 44th percentile — while still carrying the second-lowest IV rank. Cheap by the headline number, no longer thin underneath. That gap is worth watching; it’s the same shape the S&P showed two Sundays ago right before its own cushion thickened further.

The Nasdaq is the holdout. Negative VRP for a fourth straight read now — 2nd percentile three Sundays ago, 3rd two Sundays ago, 11th at last Wednesday’s check, 17th since Friday’s close. It’s the least negative it’s been in that stretch, but it survived the worst week of price action in a month without flipping positive. Realized vol has been outrunning implied on this name for over a month now.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike VIX they price this specific week, catalysts and all.

_{This week's ±1σ cones through Friday 8/28. Every cone is wider than last week's — the first time that's happened in several editions.}

Every cone is wider this week, on all three lanes — the first time that’s happened in several editions, after a run of narrowing weeks. That tracks the richer implied vol in the lane table above; wider cones are the direct consequence of paying more for the same length of time.

The calendar carries real weight this time. Wednesday brings a macro cluster — Durable Goods, PCE, and the second estimate of Q2 GDP all land the same morning, by the sources checked — and Nvidia reports that night, the single biggest name left on this earnings season’s calendar and squarely inside the Nasdaq’s now-wider cone. Then Friday morning, before this cone even settles, new Fed Chair Kevin Warsh gives his first Jackson Hole keynote as chair.

The same range, against where each index has actually been trading:

^{A month of daily candles per index, then this week's priced range fanned out from Friday's close — same ±1σ weekly expected move as the table above, not a new number.}

What it means for premium sellers

• A catalyst landing inside the pricing window is a different risk than one landing after it. Two Sundays ago, Jackson Hole sat past these cones — anyone extending exposure past that Friday was pricing a different week than the cone covered. This week the Fed Chair’s own keynote lands hours before the close these cones settle against.

• A thickening variance risk premium describes the last 20 sessions, not the next five. All three lanes are better cushioned than they were two Sundays ago, but that’s backward-looking — the S&P and Russell’s cushions widened right through a losing week last week, and there’s no mechanical reason richer implied has to keep outrunning realized into this one.

• The Nasdaq’s negative reading has now survived four straight checks across three weeks, including the worst single week of price action in a month. That’s a real pattern worth tracking rather than dismissing, but a pattern surviving one hard test isn’t the same as a rule.

• Wednesday concentrates real risk: a macro data cluster and Nvidia’s earnings, hours apart. That’s before Friday’s keynote even arrives. A week with two separate high-reach events rarely resolves as calmly as a week with one.

• I don’t know whether Friday’s keynote gets priced in gradually through the week or held to the last session. Jackson Hole speeches have moved markets on both patterns before. That’s the honest uncertainty in this issue — not a hedge, just something this cone can’t tell you in advance.

The risk lens

What a careful seller actually watches this week:

• Kevin Warsh’s Jackson Hole keynote, Friday morning, Aug 28. His first address as Fed Chair, and the same session this week’s cones settle against Friday’s close.

• Nvidia earnings, Wednesday after the close. The single biggest name left on this earnings season’s calendar, concentrated in the Nasdaq’s cone.

• Wednesday’s macro cluster — Durable Goods, PCE, and Q2 GDP (2nd estimate), all the same morning. The week’s densest single day for scheduled data, landing hours before Nvidia reports that night.

• The Nasdaq’s variance premium, negative for a fourth straight check across three weeks. It’s the least negative it’s been in that stretch — worth a fresh read next Sunday for whether that continues or reverses.

• The running cone record, 15 of 24 (62.5%), after last week’s second 0-for-3 in three weeks. Both misses broke all three names at once; worth watching whether a Fed-Chair-on-settlement-day week adds a third.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Mon 8/24 — Chicago Fed National Activity Index, 8:30 AM ET.

• Tue 8/25 — Consumer Confidence (Conference Board), 10:00 AM ET. Richmond Fed Manufacturing Index, 10:00 AM ET.

• Wed 8/26 — Durable Goods, PCE, and Q2 GDP (2nd estimate), 8:30 AM ET, all the same morning. Nvidia reports after the close.

• Thu 8/27 — Weekly jobless claims, 8:30 AM ET. Jackson Hole Economic Symposium begins.

• Fri 8/28 — New Fed Chair Kevin Warsh’s Jackson Hole keynote, Friday morning — his first as chair. This week’s cones settle against Friday’s close.

_{The same week, laid out day by day. Two top-tier days — Wednesday's cluster plus Nvidia, and Friday's keynote landing on the cone's own settlement.}

A macro cluster and Nvidia on Wednesday, then the Fed Chair’s own speech before Friday’s close. I don’t know whether this resolves as one big session or bleeds out across the week. That’s what Wednesday’s Recalibration is for.

Your turn: Every variance risk premium in this issue’s table has now thickened for three straight Sundays, even through last week’s second 0-for-3 week. Does a pattern that’s survived a losing week change how you’d size into a week with two dense catalyst days, or does the calendar override it regardless of what the cushion says? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday’s Setup called this a genuinely quiet, data-light week — no CPI, no PPI, just FOMC minutes and a retail-earnings wave — priced off two near-record-cheap IV ranks, the S&P’s at the 2.4th percentile of its own year, the Russell’s at the 0.4th. It wasn’t quiet. Renewed US-Iran tension and a bond-yield jump opened the week, a chip-sector selloff (a semiconductor gauge fell roughly 5%) hit Tuesday, and more Iran-driven pressure made Thursday the worst session of the five.

All three indexes closed Friday below the floor of Sunday’s cone. The S&P at 7,674.37 against a 7,715.9 – 7,855.6 range (−1.60σ). The Russell at 3,017.87 against 3,029.0 – 3,107.8 (−1.28σ). The Nasdaq at 29,308.86 against 29,599.4 – 30,492.8 (−1.65σ). That’s the second 0-for-3 week in this record, after August 7th’s first one. The running cone record falls to 15 of 24 — 62.5%, down from 71.4% last Friday.

Wednesday’s Recalibration flagged something strange mid-week and asked readers whether it should change how they size a premium sale: every variance risk premium was getting thicker, not thinner, during the selloff. It held all the way to Friday’s close. Implied vol outran 20-day realized on all three names across the whole week — the cushion behind the sale actually improved during the worst week of price action since early August.

The week in vol

Sunday’s issue had already flagged the widest 9-day-to-3-month gap logged in eight weeks of tracking — 7.85 points, entirely a front-end story. This week the front end did what a real shock does: it moved first, moved hardest, and then partly unwound.

• VIX9D: 10.61 → 12.58. Net +1.97, the biggest mover of the three gauges — up to 13.59 by Tuesday’s close before easing back into Friday.

• VIX: 14.25 → 15.13. Net +0.88, a smaller version of the same shape.

• VIX3M: 18.46 → 18.50. Net +0.04. The back end again barely moved, same as every week this dataset has tracked.

The front end moved first and hardest on Tuesday's chip selloff, then partly unwound into Friday. The 9-day-to-3-month gap: 7.85 Sunday, 5.68 Tuesday, 5.92 Friday.

The 9-day-to-3-month gap: 7.85 → 5.68 (Tuesday) → 5.92 (Friday). The front-end shock repriced fast and then held roughly where it landed rather than fully round-tripping back to Sunday’s stretched level. FOMC minutes printed Wednesday at 2:00 PM ET — three weeks after the meeting they documented — and showed several officials saying further rate hikes would be necessary if inflation stays elevated, a notably hawkish-leaning read for minutes this far removed from the meeting. It didn’t reverse the week’s direction on its own; Thursday’s session was worse than Wednesday’s.

All three IV ranks came off their near-record lows as implied richened: S&P 4.4 → 9.4, Russell 0.5 → 8.4, Nasdaq 21.5 → 29.3. None of them are anywhere near rich yet — all three are still bottom-third reads of their own year — but the direction is unambiguous, and it’s the same direction implied took relative to realized, which is where this week’s real story sits.

The scorecard: did the market stay inside the range?

The range settles on Friday’s close. That’s the only bell that counts for this table.

Each bar is one session's actual high-to-low range against Sunday's ±1σ cone. All three closed below the floor by Friday.

Friday’s close actually understates how far this week strayed. Thursday’s intraday low was the worst point of the week on all three — the S&P traded to roughly −2.10σ, the Russell to −2.01σ, the Nasdaq to −2.08σ — before a broad Friday bounce (major indexes added roughly 0.4–1% on the session, helped by flash PMIs and an easing in bond-market stress) pulled all three back toward, but not through, the floor.

What actually did it

Monday, oil and long-dated Treasury yields both jumped on renewed US-Iran tension, with the 30-year yield briefly touching its highest level in nearly two decades; the S&P and Dow each fell about 0.5%. Tuesday was the Nasdaq’s day: a selloff in chipmakers — a closely watched semiconductor gauge fell roughly 5% — dragged it down 1.68% on our own close-to-close read, the single worst session any index had all week. Wednesday brought the FOMC minutes described above and a modest, mixed session. Thursday was the week’s real damage for the S&P and Russell — renewed US-Iran tension and continued bond-market volatility drove the broadest selloff of the week (S&P −0.87%, Russell −1.34%, our own numbers). Friday brought flash PMIs and some relief in Treasury markets, and every index bounced, just not enough. Those are the drivers reported at the time; confirm against your own sources before leaning on the specific figures.

The running record, eight Sundays priced and eight Fridays graded:

Eight Sundays priced, graded. 15 of 24 inside — 62.5%, below what a ±1σ range is built to do.

15 of 24 — 62.5%, against a ±1σ design that expects roughly 68%. Last Friday this read 71.4%. Two 0-for-3 weeks out of eight is a lot for a design that expects a ±1σ range to hold about two-thirds of the time, and both of this record’s 0-for-3 weeks have landed within the last three. Twenty-four observations is still a small sample — worth watching whether a pattern is forming or whether this is what noise around 68% looks like at this size.

What the seller actually collected

Same-method comparison as always: what each lane charged Sunday, against what it’s actually delivered on the standard 20-day realized window.

Every ratio above 1.00× except the Nasdaq, which closed the gap from 0.87× to 0.93×. The seller was paid more than the 20-day window delivered on two of three lanes, during a losing week.

This is the week’s real finding, and Wednesday’s Recalibration only had it half-formed. Every variance risk premium widened this week, not narrowed — during a losing week for every index. The S&P’s HV20 actually fell slightly (13.30 → 13.17) even as implied rose 0.88 points; the Russell’s HV20 rose but implied rose faster; the Nasdaq’s HV20 fell too (23.70 → 22.31), pulling its cushion up from bottom-3rd-percentile to the 17th even while the index itself dropped 2.45% on the week. Selling premium into this specific selloff, on this measure, paid better by Friday than it did Sunday morning — the opposite of the intuitive direction, and it wasn’t a one-session blip; it held for the full five.

The Nasdaq’s variance premium is still negative, now a fourth straight read — 2nd percentile two Sundays ago, 3rd last Sunday, 11th at Tuesday’s mid-week check, now 17th. That’s the closest it’s come to flipping positive in a month, and it happened during the worst single week of price action any of the three indexes had in that stretch. Realized vol has been outrunning implied on this name for weeks; this week’s price damage didn’t extend that gap, it narrowed it.

The path: how much ground it took to lose 1.4–2.5%

Total distance covered versus net change. Every index covered close to four times its own priced weekly move in true range.

Efficiency is net move divided by total distance traveled — how much of the walking turned into progress. 32–43% is moderate: more directional than the chop this series has flagged as sub-30%, but with real give-back on all three, consistent with a week that opened one way (Monday/Tuesday selling), rebounded Wednesday, sold off again Thursday, then partly reversed Friday.

Every index covered close to four times its own priced weekly move in true range — S&P 3.99×, Russell 4.09×, Nasdaq 3.86×. That’s short of the 4.6× the S&P covered during this record’s first 0-for-3 week (August 7th) but well above last week’s calm 2.3–2.4×. The two 0-for-3 weeks in this record are also its two highest-travel weeks — not a coincidence a two-observation sample can prove, but not nothing either.

What it means for premium sellers

• A cone breach and a thinning cushion are two different claims, and this week they moved in opposite directions. Every index closed outside its priced range, and every variance risk premium got fatter. If your sizing logic ties directly to “did we break the range,” this week paid to be short premium on the standard 20-day measure regardless of what the scorecard says.

• This is the second time in three weeks that variance risk premium moved the “wrong” way relative to price. It isn’t yet a rule — I don’t have a clean mechanical explanation for why implied is consistently outrunning realized into these breaks rather than the reverse — but it’s now happened on two of this record’s two 0-for-3 weeks, which is worth tracking rather than dismissing as one-off noise.

• The running cone record just took its second real hit in three weeks. 15 of 24, 62.5%, against a ±1σ design built for roughly 68%. Both misses were 0-for-3 weeks, not single-index breaks — when this design fails, it has so far failed on all three names at once, which is a different risk profile than an independent per-index miss rate would suggest.

• The Nasdaq’s negative variance premium survived its worst pricing test yet. Four straight negative reads, now narrowing toward fair even as the index dropped 2.45% on the week. If this name’s realized-outrunning-implied pattern is going to break, a week like this one — real selling pressure, a sector-specific shock on top of the macro story — was a reasonable place to expect it, and it didn’t.

• I don’t know whether the front-end vol repricing (the 9-day-to-3-month gap tightening from 7.85 to 5.92) reflects a genuinely calmer week ahead or just this week’s shock working through the system. That’s the honest uncertainty here — the gap is still wider than every reading before this month, and Jackson Hole is now one week closer.

The week ahead

The full read is Sunday, so this is the shape only, and I’d rather price it off Sunday’s data than guess here.

Implied vol is higher across the board than it was a week ago — VIX 14.25 → 15.13, IV ranks of 9.4 / 8.4 / 29.3 against last Sunday’s 4.4 / 0.5 / 21.5 — while HV20 actually fell on the S&P and Nasdaq despite the week’s price action. That combination, richer implied against flat-to-lower realized, is the same VRP-thickening this issue just spent two sections on, and whether it persists into Sunday’s fresh read is genuinely open.

Jackson Hole (August 27–29) — the new Fed Chair’s first address in the role — now sits inside next week’s cone for the first time. Every issue since early August has flagged it as the real catalyst sitting one week past the current settlement. That’s no longer true starting Sunday.

Your turn: Every variance risk premium in this issue’s table finished the week thicker than it started, even as all three indexes closed outside their priced range — the same direction Wednesday’s Recalibration flagged mid-week, holding all the way to settlement. Does two data points in three weeks change how you’d read a cone breach going forward, or is that still too small a sample to act on? Reply — I read and answer every one, and the best reads make it into Sunday’s Setup.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday’s Setup called this a genuinely quiet, data-light week — no CPI, no PPI, just FOMC minutes and a retail-earnings wave — priced off two near-record-cheap IV ranks: the S&P’s at the 2.4th percentile of its own year, the Russell’s at the 0.4th. Two sessions later, quiet is not what happened.

All three indexes closed Tuesday below the floor of Sunday’s cone. The S&P at 7,691.76 against a 7,715.9 – 7,855.6 range. The Russell at 3,017.89 against 3,029.0 – 3,107.8. The Nasdaq at 29,490.96 against 29,599.4 – 30,492.8. In Sunday’s own units the S&P sits at −1.35σ.

Unlike a couple of recent breaks, this one isn’t a blowout — the distance below the floor is modest (0.31–0.37% on all three) and getting back inside by Friday costs less than half of what the market still has priced for Wednesday through Friday (0.30–0.44×). But it came at real efficiency: 70–92% of the week-to-date distance traveled converted into net move, not chop. That’s a genuine break, just a smaller one than the cone’s width might suggest.

The stranger part: every variance risk premium got thicker anyway. Implied vol richened faster than 20-day realized rose on all three names, so the cushion behind the sale actually improved during a selloff. And FOMC minutes — the week’s only scheduled top-tier catalyst — haven’t even printed yet.

Where we are vs. Sunday’s cones

Sunday’s ranges run through Friday’s close, roughly a 2-in-3 chance of finishing inside. Two of five sessions are gone. The band is Sunday’s cone, the solid bars are each session’s actual high-to-low range, the markers are the closes, and the shaded block on the left is what the market is currently charging for the three sessions that are left.

_{Sunday's ±1σ cones through Friday 8/21, with each session's actual high-to-low range drawn as a bar. All three closes are below the floor. The shaded block is the Wed–Fri move quoted from the 8/21 straddle — on all three, the distance back inside is under half of it.}

The S&P closed Monday at 7,745.06, roughly mid-cone, then gave up 0.58% Tuesday to punch through the floor. The Russell and Nasdaq both traded well inside their ranges Monday before the same Tuesday session took all three out.

Friday 8/21 is a listed expiry on all three indexes, so I can read the price of exactly the stretch that’s left rather than approximating it: ±0.72% on the S&P, ±1.11% on the Russell, ±1.22% on the Nasdaq, Wednesday through Friday. That’s the denominator in the last column, and it’s a small one relative to last month’s breaks. Getting the S&P back inside Sunday’s range by Friday’s close takes a 0.31% rally against a 0.72% move still priced — 0.44 times the entire remaining budget. The Russell needs 0.33× and the Nasdaq 0.30×.

None of that is a forecast. A cone breached Tuesday can close inside Friday, and the running record here is 15 of 21 (71.4%) across seven Sundays — a real edge for a ±1σ design expecting roughly 68%. The point is narrower: unlike the last two cone breaks logged in this series, which needed a move bigger than the entire remaining budget just to get back inside, this one is a modest overshoot that a single ordinary session could close.

What actually did it

Monday, oil and long-dated Treasury yields both jumped on renewed US–Iran tension, with the 30-year yield briefly touching its highest level in nearly two decades; the S&P and Dow each fell about 0.5%. Tuesday was sharper and more concentrated: a selloff in chipmakers — a closely watched semiconductor gauge fell roughly 5% — dragged the Nasdaq down 1.3% while elevated yields and oil weighed on the broader tape. Those are the drivers reported at the time; confirm against your own sources before leaning on the specific figures.

That maps cleanly onto the efficiency spread below: the Nasdaq, sitting on the chip selloff directly, traveled the most in absolute terms but converted the least of it into net move (70%), while the S&P and Russell — hit by the broader yield/oil story rather than a sector-specific hit — converted 90%+.

The distance, and how straight it was

Efficiency is net move divided by total distance covered — how much of the walking turned into progress. At 90%+ on the S&P and Russell there was almost no give-back: no round trip to sell into, no mean reversion to lean on. The Nasdaq’s 70% is still a real trend by the standard this series has used (anything above roughly 60–70% has read as directional, not chop), just a noisier one — consistent with a sector-specific shock (chips) layered onto the same macro pressure hitting the other two.

Two sessions is a thin window and the decimals shouldn’t be over-read. But three indexes independently landing in the 70–92% band, on the same two sessions that broke three separate cones, is a real signal, not an artifact.

What’s changed in vol

Same-method comparison — each gauge against Sunday’s read, nothing rebased.

_{VIX term structure, Sunday's read to Tuesday's close. Sunday's 7.85-point 9-day-to-3-month gap was the widest logged in eight weeks of tracking this shape; two sessions later it's 5.68, nearly all of it from the front.}

Sunday’s issue flagged the 9-day-to-3-month gap at 7.85 points as the widest we’d logged in eight weeks of tracking this shape. Two sessions later it’s 5.68 — nearly four times the 30-day gauge’s move and over three and a half times the three-month’s, concentrated almost entirely in the front end, exactly where a real near-term shock should show up first. All three IV ranks came off their near-record lows, though none of them are anywhere close to rich yet — the S&P’s 13.3 and the Russell’s 9.8 are still bottom-quintile reads.

FOMC minutes print at 2:00 PM ET today, three weeks after the meeting they document. The front-end repricing happened entirely before that — the market moving on realized oil, yield, and chip-sector news, not in anticipation of anything the Fed transcript might say this afternoon.

The cushion, despite the break

This is the counterintuitive read of the week. The market sold off, and every variance risk premium got thicker, not thinner. Implied vol on all three rose faster than 20-day realized did — the S&P’s HV20 barely moved (13.30 → 13.58) while its implied jumped over a point and a half. The Russell’s HV20 actually fell slightly even as its implied richened. Selling premium into a two-session drawdown, on this measure, paid better on Tuesday than it did Sunday.

The Nasdaq is still the one lane where the numbers disagree with the other two — negative VRP for a third straight read now, still the 11th percentile of its own year, even after improving from 3rd. It has genuinely delivered more movement over 20 sessions than its options currently charge to insure, and Tuesday’s chip-driven session is exactly the kind of move that keeps feeding that HV20 window.

The adjustment

• A cone breach on Tuesday is not a cone breach on Friday. These settle against Friday’s close, and the running record is 15 of 21 (71.4%) across seven Sundays. This week’s overshoot is also small relative to recent breaks — under half the remaining priced move on all three lanes, not the 1.2–1.8× multiples logged in early August.

• High efficiency without a big number is still a real signal. 70–92% of two sessions’ distance converted into net move. A short-premium book adjusting into this week is adjusting against a trend, even though the absolute breach is modest.

• The variance premium moving opposite the tape is worth sitting with. Every cushion in this issue’s own table got thicker during a selloff, not thinner. That is not the intuitive direction, and I don’t have a clean mechanical story for why implied outran realized here rather than the reverse.

• The Nasdaq’s negative VRP is now a three-read pattern, not a one-off. 3rd percentile two Sundays ago, 3rd again last Sunday, now 11th mid-week. It has stopped looking like noise from any single macro print.

• I don’t know whether Tuesday’s chip-sector selloff is a standalone rotation or the start of something that spreads. Nothing in this week’s data distinguishes those two stories yet, and Thursday’s retail earnings wave (Walmart, in particular) is the next real test of whether the pressure is macro-wide or sector-contained.

What’s left on the week’s calendar

• Wed 8/19 (today) — FOMC Minutes, 2:00 PM ET. July 28–29 meeting, three weeks after the fact. Target, Lowe’s, TJX, and Analog Devices report.

• Thu 8/20 — Weekly jobless claims and Philly Fed Business Index, 8:30 AM ET. Leading Indicators, 10:00 AM ET. Walmart, Alibaba, Deere, and Ross Stores report.

• Fri 8/21 — S&P Global flash PMIs (Manufacturing, Services, Composite), 9:45 AM ET. This week’s cones settle against Friday’s close.

_{What's left on the week's calendar. FOMC minutes this afternoon are the only scheduled top-tier item left; the rest is the retail-earnings wave. Jackson Hole (Aug 27–29) still sits outside next week's cone too.}

Jackson Hole (Aug 27–29) — the new Fed Chair’s first address in the role — still sits outside next week’s cone as well as this one. Nothing on the calendar between now and Friday changes that; the settlement Friday will be judged on a week where the real scheduled catalyst still hasn’t arrived.

Your turn: Every variance risk premium in this issue’s table got thicker during a two-session selloff, not thinner — implied outran realized on all three indexes. Does that change how you’d size a premium sale into a break like this one, or does the direction of the underlying just not matter to that decision? Reply — I read and answer every one, and the best reads make Friday’s Debrief.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

The S&P’s implied vol is priced at the 2.4th percentile of its own past year (IV rank 4.4). The Russell’s is at the 0.4th (IV rank 0.5). Those are real numbers off a real 252-session window — genuinely rare reads, no extra framing needed to make that point.

What complicates it: HV20 has kept climbing under both names, so the cushion behind the headline number isn’t quite as thin as “2nd percentile” makes it sound — the S&P’s variance premium actually ticked up slightly this week (7th percentile → 11th), and the Russell’s ticked down a touch (36th → 31st). The Nasdaq is the one lane where cheap and thin agree: its VRP is negative for a second straight Sunday, 3rd percentile, meaning it’s delivered more movement over 20 sessions than its options currently charge to insure.

The calendar backs up the “quiet” read — this is a genuinely data-light week: no CPI, no PPI, just FOMC minutes Wednesday and a wall of retail earnings. The thing that isn’t quiet sits one week past where these cones settle: Jackson Hole, August 27–29, where new Fed Chair Kevin Warsh gives his first address in the role.

The regime right now

Last week’s issue flagged records on the board and a Nasdaq VRP gone negative for a fourth session. This week both indexes got cheaper still, and the Nasdaq stayed negative — nothing snapped back.

• VIX 14.25 — down 0.65 from last Sunday’s 14.90. Lower half of normal.

• Term structure: 10.61 → 14.25 → 18.46 (9-day → 30-day → 3-month). Still contango. The 9-day-to-3-month gap is now 7.85 points, the widest we’ve logged in the ~7 weeks we’ve been tracking this shape — worth noting for what it’s worth, not leaning on as a historical extreme; it’s a small sample. The front did nearly all the moving again: VIX9D down 1.35, VIX3M down just 0.26.

The curve, last Sunday vs. now. The gap widened again, 6.76 → 7.85 — the widest we've logged yet.

The receipt from last week: the S&P closed net +0.36% on 2.84% of travel (13% efficiency), the Russell net +1.12% on 3.42% (33%), the Nasdaq net +1.09% on 4.99% (22%). Low efficiency across the board — the tape drifted more than it trended, which is exactly the kind of week that lets HV20 rise without implied following it up.

Last 21 sessions per index, cumulative % change, against realized-vol bands built from each index's own HV20 — backward-looking, not a forecast. Last 5 sessions shaded.

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

Variance risk premium, prior Sunday vs. now, ranked against each index's own past year. The S&P's cushion widened even as its headline vol fell further — the split this week's TL;DR flags.

A few things worth reading off this slowly.

The S&P’s headline number and its cushion just disagreed for the first time in a while. IV rank fell (8.1 → 4.4) but the variance premium rose (7th percentile → 11th) — HV20 fell faster than implied did. Cheap on the headline doesn’t automatically mean thin underneath; this week it means the opposite, slightly.

The Russell is cheap by both measures, but not extreme by both. IV rank at the 0.4th percentile of its own year is about as low as this gauge goes. Its variance premium sits at the 31st percentile — moderately paid, not razor-thin. Same split as last week, just a shade less pronounced.

The Nasdaq has barely moved. Negative VRP two Sundays running, 2nd percentile last Sunday to 3rd now — essentially flat, not identical. That’s now a real week of realized vol consistently outrunning what its options charge — worth watching whether it persists into a data-light week, since there’s less on the calendar to explain it away this time.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike VIX they price this specific week, catalysts and all.

This week's ±1σ cones through Friday 8/21. Every cone narrowed again; the calendar is data-light — FOMC minutes Wednesday, a retail-earnings wave the concrete risk.

Every cone narrowed again this week, on all three lanes — a continuation of the pattern the last several issues have flagged. That’s a real, plainly-observable move in the numbers above, not a claim about where it ranks against any longer history we don’t have.

The structural shape is different from the last two weeks. No CPI, no PPI, no jobs report inside this window — the only scheduled catalyst with real reach is Wednesday’s FOMC minutes, and minutes from a meeting three weeks gone tend to move markets less than the meeting itself did. The retail-earnings wave (Home Depot, Target, Lowe’s, TJX, Walmart, and others) is the more concrete risk, concentrated in the Nasdaq and small-cap names more than the S&P.

The same range, against where each index has actually been trading:

A month of daily candles per index, then this week's priced range fanned out from Friday's close — same ±1σ weekly expected move as the table above, not a new number.

What it means for premium sellers

Educational framing, not a trade — conditions, not signals.

• A rare percentile and a thin cushion are two different claims, and this week they don’t move together on two of three lanes. The S&P’s headline vol is genuinely cheap by any real measure, but its variance premium actually widened slightly. Check both before assuming “cheap IV rank” means “thin cushion.”

• A negative variance premium describes the last 20 sessions, not the next five. The Nasdaq has delivered more than it’s charged for across two Sundays now. That’s a real pattern worth tracking, but it isn’t a forecast for this specific data-light week.

• A data-light week is not the same as a risk-free week. No CPI or PPI doesn’t mean nothing can move the tape — FOMC minutes can surprise, and a wall of retail earnings carries real single-name and sector risk even without a top-tier macro print behind it.

• The real catalyst sits outside this week’s cone entirely. Jackson Hole (Aug 27–29) and the new Fed Chair’s first address as chair (Aug 28) both land after these options settle. Anyone extending exposure past Friday is implicitly pricing a different week than the one this cone covers.

• I don’t know whether “quiet” holds through Jackson Hole, or whether that’s exactly the kind of setup that gets unwound in one session once a real catalyst shows up. I don’t have a clean answer either way — it’s the honest uncertainty in this issue, not a hedge.

The risk lens

What a careful seller actually watches this week:

• FOMC minutes, Wednesday 2:00 PM ET. The only top-tier scheduled macro event this week; minutes from the July 28–29 meeting, three weeks after the fact.

• Home Depot, Tuesday before the open. First of the week’s retail-earnings wave.

• Target, Lowe’s, TJX, and Analog Devices, Wednesday. The bulk of the retail print lands the same day as the FOMC minutes.

• Walmart, Alibaba, Deere, and Ross Stores, Thursday. The wave’s biggest single name (Walmart) closes it out.

• The Nasdaq’s variance premium, negative for a second straight Sunday. Worth a fresh read next Sunday — a third negative reading in a data-light week would be a cleaner signal than the last two, which both had a macro print to point to.

• Jackson Hole, Aug 27–29 — outside this week, but on the calendar. Not this week’s risk, but worth knowing it’s the next scheduled node with real reach.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Mon 8/17 — Empire State Manufacturing Index, 8:30 AM ET. NAHB Housing Market Index, 10:00 AM ET.

• Tue 8/18 — Building Permits and Housing Starts, 8:30 AM ET. Industrial Production, 9:15 AM ET. Home Depot reports before the open.

• Wed 8/19 — FOMC Minutes, 2:00 PM ET. Target, Lowe’s, TJX, and Analog Devices report.

• Thu 8/20 — Weekly jobless claims and Philly Fed Business Index, 8:30 AM ET. Leading Indicators, 10:00 AM ET. Walmart, Alibaba, Deere, and Ross Stores report.

• Fri 8/21 — S&P Global flash PMIs (Manufacturing, Services, Composite), 9:45 AM ET. This week’s cones settle against Friday’s close.

Two genuinely rare vol reads, a data-light week to sit through, and the actual test one week past where these cones close. I don’t know if quiet stays quiet. That’s what Wednesday’s Recalibration is for.

Your turn: The S&P’s headline vol rank fell this week but its variance premium actually widened slightly — cheap and thin split apart. Does that change how you’d read “2nd percentile” on its own, or is a rare vol reading a rare vol reading regardless of what the cushion underneath is doing? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday’s issue flagged two records: the S&P’s IV rank at 8.1, the Russell’s at 1.8, both the cheapest either has read since this dataset started tracking in late June. Then the week actually delivered the test those records were supposed to face — CPI Wednesday, PPI Thursday, and a weaker-than-expected retail sales print Friday morning, three real data points in one week, not zero.

All three cooled. CPI came in at 3.4% year-over-year, down a tenth from June; PPI was flat against a 0.2% consensus; retail sales fell 0.6%, the sharpest drop since last May. Normally that’s when a record-cheap reading snaps back — the market got its news, and the news didn’t shake anything. It didn’t snap back. It went cheaper. S&P IV rank: 4.4. Russell: 0.5. Both new all-time lows, one week after the old ones.

The scorecard agrees with the calm: all three indexes closed inside Sunday’s priced range (S&P +0.31σ, Russell +0.75σ, Nasdaq +0.51σ) — the first 3-for-3 week since July 26 and a full reversal of last Friday’s first-ever 0-for-3.

The week in vol

CPI and PPI both cooled, and the vol complex didn’t treat that as news worth pricing. It treated it as one more reason to get cheaper.

• VIX9D: 11.96 → 10.61. Net −1.35, the biggest mover of the three gauges. It actually rose into Monday and Tuesday (12.77, 12.52) before CPI, then fell hard Wednesday and never recovered.

• VIX: 14.90 → 14.25. Net −0.65, a smaller version of the same shape — up into the print, down after it.

• VIX3M: 18.72 → 18.46. Net −0.26. The back end barely moved all week, same as every week this dataset has tracked.

CPI cooled Wednesday and the front end kept falling anyway. The 9-day-to-3-month gap closed the week at a record 7.85.

The 9-day-to-3-month gap set a new record: 7.85, beating the old high of 7.42 from July 12 — which Sunday’s issue had already flagged as the second-widest reading on the board at 6.76. The gap actually crossed the old record on CPI day itself (7.44 at Wednesday’s close, ahead of the prior high) and kept widening into Friday. This is now the fourth straight week the gap has moved, and the third straight week it’s widened.

None of the three single-name index gauges moved more than a point and a half in either direction all week. Whatever else happened in the tape, the options market read CPI, PPI, and a soft retail sales number as confirmation, not surprise.

The scorecard: did the market stay inside the range?

The range settles on Friday’s close. That’s the only bell that counts for this table.

The range settles on Friday's close. All three finished inside it — the first 3-for-3 week since 7/26.

Each bar is one session's actual high-to-low range inside the week's priced range. All three stayed inside all week.

None of the three came close to a bound this week: the Russell’s +0.75σ is the closest anything got, and that’s still a quarter of a standard deviation from the ceiling. Every close and every intraday session sat inside Sunday’s shaded band, which is a genuinely different shape from the last two Fridays — 8/7 broke every ceiling, 7/31 held every range too. This week is closer to 7/31 than to last week, the calmer pattern, not the volatile one.

The running record, seven Sundays priced and seven Fridays graded:

Seven Sundays priced, graded. 15 of 21 inside — 71.4%, back above what a ±1σ range is built to do.

15 of 21, 71.4%, against a ±1σ design that expects roughly 68%. Last Friday this read 66.7%. One 3-for-3 week moved it five points on a denominator of 21, the same lesson last week’s table taught in the other direction. Twenty-one observations is still not a sample to build a rule on.

What the seller actually collected

Same-method comparison as the Setup’s lane table: IV rank last Sunday vs. now, plus what each index actually delivered.

Both records that Sunday flagged fell further. The Nasdaq's variance premium is unchanged to the second decimal — still negative, still the worst-paid lane.

The S&P’s and Russell’s records didn’t just hold, they broke again. IV rank 8.1 → 4.4 on the S&P, 1.8 → 0.5 on the Russell. HV20 fell on all three this week (S&P 14.4 → 13.3, Russell 15.7 → 15.4, Nasdaq 26.1 → 23.7), so the variance premium barely moved even as the headline rank cratered: S&P 1.04× → 1.07×, Russell 1.18× → 1.15×. Cheap and thin are usually the same story; this week they mostly agree, just at an even cheaper level than Sunday found.

The Nasdaq’s variance premium is still negative — 0.87×, now unchanged from Sunday’s read to the second decimal, still the worst-paid lane on the board and still bottom-3rd-percentile of its own year. Six sessions running now (the four Sunday counted, plus this week), the standard 20-day window keeps saying the index has delivered more than its options currently charge.

Total distance covered versus net change. Every index covered more than double its own priced weekly move in true range.

Distance traveled tells a different story than the calm scorecard does. Every index covered more than double its own weekly cone in true range: the S&P 2.84% against a ±1.17% cone (2.4×), the Russell 3.42% against ±1.50% (2.3×), the Nasdaq 4.99% against ±2.15% (2.3×). Efficiency (net move divided by distance traveled) came in low across the board: 13% on the S&P, 33% on the Russell, 22% on the Nasdaq, meaning most of that ground was round-tripped, not delivered. Wednesday’s Recalibration read 36% / 18% / 43% off just two sessions; three more sessions of chop pulled the S&P’s and Nasdaq’s efficiency down further while the Russell’s actually improved. A book that was short gamma paid for that distance regardless of what the closing prices eventually agreed to say.

What it means for premium sellers

• A record is a description of the past, not a forecast about the news. Sunday’s issue said exactly that about the 8.1 and 1.8 readings, and CPI, PPI, and retail sales all landing without incident this week is the closest thing to a direct test those words have gotten. The record didn’t snap back to test how firm it was — it moved to a new low. That’s one data point about how these records behave, not a rule.

• Cheap headline rank and thin variance premium are now the same claim on two of three lanes. The S&P and the Russell both have record-low IV rank and a variance premium sitting in the bottom third of their own year. There’s no gap left to lean on for either one — you’re being paid close to the least you’ve been paid all year, on indexes that still covered 2.3–2.4× their own weekly cone in true range.

• The Nasdaq’s negative variance premium is now a multi-week fact, not a one-day anomaly. Six sessions of realized outrunning implied on the standard 20-day window. That’s worth checking again after the next macro print, the same caution Sunday’s issue carried forward.

• Travel and the scorecard told different stories again. A reader who only checked Friday’s closing prices would call this the calmest week of the last three. A reader who checked true range would see every index cover more than double its priced weekly move to get there. Both readings are correct; neither one is the whole picture on its own.

• Three real data points, one direction. CPI, PPI, and retail sales all landed this week, and vol treated all three the same way — as confirmation the current regime holds, not as reasons to reprice. Three prints agreeing is worth more than any one of them alone, and it’s still three prints, not a law.

The week ahead

The full read is Sunday, so this is the shape only, and I’d rather price it off Sunday’s data than guess here.

Implied vol is lower across the board than it was a week ago — VIX 14.90 → 14.25, and IV ranks of 4.4 / 0.5 / 21.5 against last Sunday’s 8.1 / 1.8 / 36.2. HV20 fell on all three this week too, the first time that’s happened after three straight Sundays of realized vol rising into a cheap headline reading. I don’t know whether implied and realized falling together changes what next week’s cones look like, or whether it’s just this week’s version of the same cheap regime continuing. That’s a Sunday question, not a Friday one.

The retail sales miss Friday morning is the one loose thread: a real downside surprise on the economy, absorbed by the market on the same day these ranges settled without moving a single gauge more than a point. I’d have expected that to show up somewhere in the vol complex and it didn’t, at least not yet.

Your turn: Two records got tested by real news this week — CPI, PPI, and retail sales all landed, and instead of snapping back, the S&P’s and Russell’s IV rank both fell to new lows. Does a record that survives a real test change how you’d size around it, or does “cheap” stop mattering once it’s this cheap regardless of what’s tested it? Reply — I read and answer every one, and the best reads make it into Sunday’s Setup.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

It’s been a quiet two sessions. The S&P closed Tuesday net −0.38% since Friday, the Russell −0.24%, both within half a percent of where the week started. The Nasdaq drifted a bit further, −0.66%, but none of the three has moved enough on its own to say much. CPI prints Wednesday 8:30 AM ET, PPI Thursday 8:30 AM ET, the two prints this week was actually built around, and neither has happened yet.

Underneath those flat closes, though, there was real movement. The S&P and the Russell each covered roughly seven-eighths of their entire week’s priced move in true range: 1.04% against a 1.17% cone, 1.32% against 1.50%, at 18–36% efficiency, meaning most of that distance round-tripped rather than went anywhere. That’s chop, not a trend, and it’s the reason a scoreboard that reads “barely moved” isn’t the same thing as a week that’s actually been calm to hold.

Sunday’s other numbers mostly held too. The S&P’s and Russell’s IV ranks are still close to the record lows Sunday found (8.1 → 10.3, 1.8 → 5.7), and the Nasdaq’s variance premium is still negative, still the worst-paid lane on the board. None of this is a read on Wednesday. It’s a description of a market that hasn’t gotten its actual news yet.

Where we are vs. Sunday’s cones

Sunday’s ranges run through Friday’s close, roughly a 2-in-3 chance of finishing inside. Two of five sessions are gone. The band is Sunday’s cone, the solid bars are each session’s actual high-to-low range, the markers are the closes, and the pale block on the right is what the market is currently charging for the three sessions that are left.

Sunday's ±1σ cones through Friday 8/14, with each session's actual high-to-low range drawn as a bar. Every close is inside, most of them close to where the week started — but the room to the nearer bound, quoted from the actual Wed–Fri straddle, is already thin on the S&P and the Russell.

Every close sits comfortably inside Sunday’s range. The intraday lows don’t sit quite as comfortably — all three traded down to roughly −0.44σ to −0.47σ at some point across Monday and Tuesday before closing back up toward −0.2σ to −0.3σ. That’s not a breach; it’s a reminder that a close is a single sample of a session that moved more than the close shows, which is exactly what the travel figures below are built to catch.

The Russell is the one that actually diverges session to session: it closed Monday at −0.38σ and recovered to −0.16σ by Tuesday, while the S&P and Nasdaq both drifted the other way (S&P −0.05σ → −0.32σ, Nasdaq −0.16σ → −0.31σ). Three sessions remain, and Friday 8/14 is the actual expiry these cones settle against, so I can read the price of exactly the stretch that’s left rather than inferring it: ±0.93% on the S&P, ±1.41% on the Russell, ±1.40% on the Nasdaq, Wednesday through Friday. That’s the denominator behind every cushion figure in this issue.

Quiet at the close. Not quiet underneath.

Ground covered in two sessions against the cone drawn for five, net move marked. All three indexes have already spent 71–89% of a full week's priced move, at 18–43% efficiency — chop, not trend — while netting almost nothing, before CPI or PPI have printed.

Efficiency is net move divided by total distance covered — how much of the walking turned into progress. The Russell’s 18% is the clearest chop on the board: 1.32% of true range covered against a net move of −0.24%, most of it round-tripped. The Nasdaq’s 43% is the highest of the three and still means well over half the distance it covered came back out.

Read the two tables together and the shape is: the closes say quiet, the true range says otherwise, and both are correct at once. That’s the opposite of Ed. 5’s break two weeks ago, where the S&P ran 89% efficiency in one direction and the closes told the real story on their own. This week a reader who only checked the closing price would have a fair picture of where things ended up and a genuinely misleading picture of what it took to get there.

What changed in vol

Same-method comparison — each gauge against Sunday’s read, nothing rebased.

The nine-day gauge did essentially all the moving: three times the thirty-day’s change, nearly three times the three-month’s. Sunday’s gap was the second-widest reading since this dataset started tracking; two sessions later it’s narrowed by about a half point, entirely from the front. That’s a small unwind of a record-wide contango, not a real repricing of anything past this week, since the 30-day and 3-month gauges barely moved.

Both IV ranks that hit records Sunday ticked up off their floors (S&P 8.1 → 10.3, Russell 1.8 → 5.7), which is closer to what records tend to do than what a genuine regime shift looks like. Neither is close to leaving record-cheap territory.

The cushion Sunday flagged, and where it actually sits

Sunday’s lane table put the S&P’s variance premium at the 7th percentile of its own year and the Russell’s at the 36th — cheap headline rank, thin cushion on one, less thin on the other. The Nasdaq’s premium was negative for a fourth straight session.

All three moved the same direction, slightly better, and the mechanism is mostly HV20, not IV. Realized 20-day vol fell on all three (S&P 14.37 → 14.05, Russell 15.65 → 15.45, Nasdaq 26.1 → 24.9) as the violent stretch two weeks ago starts rolling out of the twenty-session window, while implied barely moved. The Nasdaq is still negative, still charging less than it has delivered over the past twenty sessions, still the worst-paid lane on the board. Two ordinary sessions haven’t changed that read; they’ve made it marginally less bad.

None of this is a forecast about Wednesday. It’s a description of where the cushion actually is heading into it, which two sessions ago was a Sunday estimate and is now a quoted market price.

The adjustment

• A quiet close is not the same claim as a quiet market. 18–43% efficiency across two sessions means the tape has been paying in true range without paying out in direction. A book that’s been getting whipped on adjustments has already spent real cost getting nowhere — that cost doesn’t show up in the closes.

• CPI and PPI are what this issue is actually about, and neither has happened. Everything above describes a market that hasn’t yet digested the two prints this week was built around. Wednesday 8:30 is the first test.

• The room left to Sunday’s floor is thin on two of three lanes, even though nothing has broken. Quoted off the actual Friday straddle, an ordinary-sized remaining move reaches the floor on the S&P (0.85×) and the Russell (0.89×) — arithmetic, not a forecast. An ordinary session could close that gap. An ordinary session just as easily doesn’t.

• The Nasdaq’s premium is still negative. 22.38 charged against 24.90 delivered on the standard window. That’s the fourth-plus session this has been true, not a new one-day read.

• Map the expiry to Wednesday and Thursday mornings specifically. Anything expiring before Wednesday’s open has already avoided CPI; anything expiring after Thursday’s open is short both prints. That’s a different exposure than “three sessions left” implies on its own.

I don’t know whether the front-end vol move this week, modest and mostly narrowing, means the options market expects an orderly CPI or just hasn’t gotten around to pricing it yet. Wednesday morning settles that one directly.

What’s left on the week’s calendar

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Wed 8/12 — CPI, 8:30 AM ET. Cisco reports after the close.

• Thu 8/13 — PPI, 8:30 AM ET. Weekly jobless claims the same morning.

• Fri 8/14 — Retail Sales, 8:30 AM ET. This week’s cones settle against Friday’s close.

Friday’s Debrief writes the scorecard — including whether two genuinely quiet sessions told you anything about how the rest of the week went.

Your turn: Two sessions of near-flat closes, and neither CPI nor PPI has printed yet. Does a quiet start change how you’d think about sizing into Wednesday’s print, or is “quiet so far” just what a market looks like in the hours before news it hasn’t priced in yet? Reply and I’ll read and answer every one; the best reads make Friday’s Debrief.

XOXO,

Theta Desk

Subscribe now

You’re short a put, it’s 0DTE, and the index just traded through your strike. Screen goes red, and some part of you starts negotiating: hold, or get out.

This is the first entry in a new series I’m calling 0DTE SPX Studies, and it works differently from the studies before it. Every prior study in this newsletter has priced options off a model or the published VIX band, applied to daily closes, going back twenty years — useful for what’s been true for decades, but blind to anything that happens inside a trading day. This series trades that history for precision: real SPXW minute-by-minute quotes, four years instead of twenty, so I can finally answer questions about what actually happens between the open and the close, not just where the index landed at the end of it. The touch question above is the first one worth asking with data like that.

I wanted to know what that moment actually predicts, so I went and measured it: every 0DTE SPXW session for four years, minute by minute, at five different distances from the money. Every position in this piece opens at 09:35 ET, five minutes after the bell — a touch only counts if the strike trades through after that moment, since the position isn’t open yet before it. Did the strike ever trade through after entry? If it did, did the trade still finish out of the money anyway?

The honest answer isn’t the reassuring one I expected going in. But it explains something that’s been sitting unexplained in this series since Study #3: why a strategy that wins 88% of the time can still net to roughly nothing. And once I had the put side built, I ran the same measurement on calls — 0DTE sellers run both sides of the chain, and the comparison turned out to matter more than I expected.

Subscribe now

How often does it even happen

Touch rate vs. what actually finished in the money — not the model's nominal delta.

Start with the basic fact, because it’s worth knowing on its own: a short strike gets touched roughly twice as often as it finishes in the money.

A 15-delta put — nominally “15% chance of finishing in the money” — gets touched 23% of the time. A 25-delta put finishes ITM about 22% of the time but gets touched 41%. Even a far-out 5-delta put, sold 75 points from the money, trades through 1 session in 18.

None of that is a flaw in delta as a concept — delta was never promising you a quiet ride, just a probability at expiry. But it means the false alarms alone are common. Most traders who’ve sold 0DTE premium have watched their strike trade through and survived to collect the credit anyway. The question is what that moment was actually telling them.

Once it’s touched, it’s close to a coin flip

Recovery given a touch, by delta. The dashed line is 50%.

I went in expecting the answer to be reassuring: something like “most touches are just noise, don’t panic.” That’s not really what the data says.

Once a strike is touched, whether it goes on to settle out of the money sits at 46-50% across every delta from 10 to 35. That’s a coin flip, not a “usually fine.” The one exception is the far wing: a 5-delta put that gets touched at all recovers only 38% of the time. When a strike that far away trades through, it’s disproportionately a real move, not a wobble.

There’s a clean reason the number lands so close to 50%, and it’s worth knowing even though it isn’t an edge: for a market with no strong intraday drift, touching a level and then being roughly equally likely to end up on either side of it is close to what basic random-walk math predicts. I’m not claiming SPX is a textbook random walk, it isn’t quite, but the data is landing close enough to that null that I don’t think there’s a clean “stay calm” rule hiding in the recovery rate by itself.

So a touch is genuinely uncertain. What decides the outcome isn’t whether it recovers. It’s what recovering and not recovering are each worth.

Does it matter WHEN the touch happens?

The random-walk read makes a real, testable prediction: if it’s right, recovery shouldn’t care whether the touch happens at 9:40 or at 2 PM. I split touched sessions into early/middle/late thirds by how long after entry the touch occurred, at each delta with enough sample to trust:

Flat, within noise, at both. That’s the prediction holding up, not just asserted. The 15-delta cut has a smaller sample (240 touched, ~80 per bucket) and bounces more: 40% / 58% / 41%. But it isn’t a monotone trend either, and I’d want a lot more data before reading anything into it. I checked, and I don’t think there’s a clock-based edge hiding here.

Touch rate itself is also stable across years I’d expect to behave very differently. At 15-delta it ran 22-25% every single year from 2022 through 2026, a bear market and two real shocks included. Recovery bounces year to year (36% one year, 53% another), which is exactly what you’d expect from a true rate near 46% measured on 30-60 touches a year, not evidence the mechanism itself is shifting.

The actual mechanism — and it explains the win-rate illusion

Hold-to-settlement P&L, net of real fills, split by whether the strike was ever touched.

This is the finding I didn't go looking for. I split every 09:35 short-put credit spread by whether it was ever touched, and looked at what holding to settlement actually paid, net of real execution costs:

The strategy doesn’t lose because it loses often — Study #3 already showed the win rate sits near 88%. It loses because the untouched majority collects a small, boring, reliable credit, and the touched minority hands back a multiple of it, except at the far wing, where that multiple gets extreme. A touched 5-delta put loses on average 22 times what an untouched one makes, because the credit was tiny to begin with and the spread still has the same width to lose.

Notice how fast that ratio collapses as you move in. By 25-delta, a touch costs only 1.5x the untouched win. By 35-delta, it’s actually smaller — the untouched side is now paying enough that a touch doesn’t even erase it on average, only on the worst instances. Selling closer to the money doesn’t make touched sessions painless; it makes the untouched sessions carry enough weight to matter against them. That’s the same flat-per-unit-of-risk shape Study #2 found on a completely different dataset, showing up again from the touch side.

I checked this three ways before trusting it: the strikes in this table match the ones already published in the stop-loss study exactly, zero mismatches; every untouched session settled at or above its strike, with no exceptions, which is what the definition requires; and the touched/untouched averages recombine, weighted by how often each happens, back to the exact hold-to-settlement number already on the record. It reconciles.

Does this hold for calls too?

Everything so far is the put side — a trader short volatility on the downside. But 0DTE sellers run both sides of the chain: short calls, iron condors, the works. So I built the exact same measurement on calls — same delta ladder, same 09:35 entry, same four years of quotes, same 10-wide credit spread, priced through the same fill model. The comparison isn’t a rerun of the put story with the sign flipped. It’s a different, more interesting result.

Start with the thing that makes puts and calls genuinely different: at the same nominal delta, a put sits farther from the money than a call. A 15-delta put sells 43 points from the forward; a 15-delta call at the same moment sells only 37 points away. That gap holds at every rung of the ladder — it's the market's skew, priced directly into distance. SPX has always priced crash risk richer than melt-up risk; this is what that actually looks like in points, not just in implied vol.

Here's the part I didn't expect. Given that calls touch more often, you'd guess the whole barbell looks worse for calls, but it doesn't, not in the way that matters. Once a call actually gets touched, the dollar loss is nearly identical to a touched put at the same delta: −$315 for a touched 15-delta call versus −$317 for a touched put; −$267 versus −$245 at 25-delta; even at 5-delta, where the put loss looks extreme at −$450, the call loses "only" −$326, still large, still the same barbell shape, just not as stretched. The mechanism itself (small reliable win, occasional sharp loss) is not a put-specific phenomenon. It's what happens on both sides of a 0DTE credit spread once you're on the wrong side of the strike. What skew buys you is a lower chance of getting there in the first place, not a softer landing once you do.

One more piece, held to a lower standard of confidence than the rest of this study: given a touch, calls recover less reliably than puts — 39-42% across the ladder, versus 46-50% for puts. My read is that this is the sample’s own realized drift talking, not a permanent property of calls versus puts: SPX trended up across most of 2022-2026, so a touched call was disproportionately likely to be touched because the market kept running, and a touched put disproportionately likely to be touched on a move that then partially reversed. I’d want to split this by up-years and down-years before I’d stand behind it as a structural claim, and I haven’t done that work yet. Reported as observed, not as settled.

How far past the line it actually goes

Touched sessions only. Most touches are shallow; a real minority run much deeper.

For the sessions that do get touched, how deep does the breach usually run? A typical touch carries the index 15-19 points past the strike at its worst point, not a graze and not a collapse. The distribution has a real tail, though: some touches barely cross the line and drift back, others run considerably further, and that spread is exactly what’s driving the touched-side loss in the table above.

Does the entry time change any of this

All five deltas, all four entry times. The ladder shape holds throughout.

I checked this at four entry times — 09:35, 10:00, 10:30, 11:00 — because the whole shape of this study would be a lot less interesting if it only held at one moment of the day. It holds. Touch rate eases modestly the later you get in (less session left for it to happen), and the barbell shape — small reliable win, occasional severe loss — shows up at every single entry point I tested. This isn’t a study about timing, though, and I’m not going to pretend a robustness check is one. Whether a different entry time actually changes the P&L, net of a real exit strategy, is its own question. I’m building that one next.

What I’d actually take from this

A touch is not a verdict, and it’s not noise either. It’s real, uncomfortable information that you’re now in a minority bucket where the downside dwarfs the upside you were originally paid. Knowing that doesn’t tell you what to do about it — that’s a different, harder question, and I’m going to dedicate a future study to it.

Skew buys you fewer touches, not smaller losses when they happen. If you’re deciding between structuring risk on the put side or the call side, or running both, as most 0DTE iron condor sellers do, the dollar cost of getting it wrong is not the lever skew is pulling. How often you get there is.

The 88% win rate was never lying about its number. It was lying about its meaning. A strategy can win almost 9 times in 10 and still net to nothing, if the 1-in-10 (or in this case, closer to 1-in-4 at the touch level) is big enough. If you only track win rate, you are tracking the part of this trade that matters least.

There’s further research needed around what is the correct action to take at the moment of a touch. Obviously this is dependent on your trading strategy and risk tolerance, but more data here would be helpful. This is a planned future study, stay tuned for this.

What’s next

This was my first study in what will be a recurring series on selling premium on short dated SPX options. I’m hoping to provide some real insight to the community on how to be profitable selecting and managing these kinds of trades, using real options data priced down to the minute.

Subscribe to keep receiving more detailed 0DTE studies, and let me know if there any specific studies you’d like me to run, I’d love to hear your ideas.

Subscribe now

TL;DR

Two records this morning. The S&P’s IV rank is 8.1 — the lowest read since this dataset started in late June. The Russell’s is 1.8, also a new low, and the percentile under it, 0.4, means RVX has traded above today’s level essentially the entire past year.

Normally that’s the whole issue: a big rally emptied out near-term fear, options got cheap, move on. What complicates it: realized vol didn’t cooperate. HV20 rose on all three indexes for a third straight Sunday, so every variance premium fell too, and the Nasdaq’s has now been negative for four sessions running — 0.874×, the worst of the four, meaning the index has delivered more movement than its own options have charged for since Tuesday.

Then there’s the calendar. CPI prints Wednesday, PPI Thursday — a genuine inflation double-header, landing squarely inside a week where two of the three lanes are pricing about as little as they’re capable of pricing.

Subscribe now

The regime right now

Last Sunday’s issue flagged a bottom-quartile cushion on two of three lanes heading into Friday’s jobs report. Whatever that report said, the tape read it as good news: all three indexes ripped higher into it and kept going — I don’t have a clean read on why beyond “the market liked it,” and I’m not going to pretend otherwise.

• VIX 14.90 — down 1.09 from last Sunday’s 15.99. Lower half of normal, and closing in on levels this dataset hasn’t seen.

• Term structure: 11.96 → 14.90 → 18.72 (9-day → 30-day → 3-month). Still contango. The 9-day-to-3-month gap widened to 6.76 points, the second-widest reading since tracking began — behind only the 7.42 we saw on July 12. The front did almost all of the moving again: VIX9D down 1.09, VIX3M down just 0.30.

The curve, last Sunday vs. now. The gap widened again, 5.97 → 6.76 — second only to the 7.42 read on July 12.

The S&P’s IV rank and the width of this gap both hit records today, and both trace back to the same shape the curve had on July 12. We’re re-testing the cheapest week of the whole dataset.

The receipt from last week, since we keep those: the S&P closed the week net +3.58% on 6.00% of travel, 60% efficiency — an unusually clean, directional week. The Nasdaq covered 10.61% of ground to net +5.12% (48% efficiency), the Russell 7.32% to net +3.52% (48%). That’s the realized surge now sitting inside the HV20 numbers below, which is exactly why implied fell and realized rose at the same time.

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

Every cushion on the board compressed again. The S&P and Nasdaq both sit in the bottom decile of their own year; the Russell's is thinner but not extreme.

A few things worth reading off this slowly.

The S&P’s two numbers finally agree, and not in a good way. Cheapest IV rank on record, and its variance premium fell right alongside it — 25th percentile last Sunday, 7th now. When the headline number and the cushion both hit lows together, there’s no gap left to lean on. You’re being paid the least you’ve been paid all year, for an index that’s still moving.

The Russell is the interesting exception. Its IV rank fell further than the S&P’s, in relative terms — 0.4th percentile of its own year, essentially an all-time cheap read for RVX. But its variance premium sits at the 36th percentile, not the single digits. Cheap in absolute rank terms and still moderately paid against what it’s actually delivered are two different claims, and this week the Russell is the one lane where they don’t move together.

The Nasdaq’s is the one I keep checking. Negative VRP means realized has out-run implied — the index moved more over 20 sessions than its options currently charge to insure. This is the fourth session in a row it’s been true, and Friday’s read is the worst of the four. I called this out three Sundays ago as a 10-day-window anomaly worth watching. Four sessions later, it’s showing up on the standard 20-day window too.

How that HV20 number actually happened

The lane table above gives you HV20 as one number per index. Here’s what the month behind that number actually looked like — the last 21 closes per index, plotted as cumulative % change, against bands built from each index’s own realized vol. This is the same lens as the lane table, just drawn through time instead of collapsed to a single figure — not a support/resistance chart, no trendline, no price target.

Last 21 sessions per index, cumulative % change from 7/10, against realized-vol bands built from each index's own HV20 — backward-looking, not a forecast.

Two things stand out. The Russell never left its own ±1σ band all month, and the calmest lane on IV rank was also the most orderly path, no single session doing outsized work. The Nasdaq is the opposite story. It broke its own band on July 29, closing that session down 8.83% from the start of the window against a ±7.35% line, then round-tripped almost the whole way back and landed at −0.34% by Friday. That’s the same fact as the negative VRP above, seen from a different angle: a month built from one outsized drop and a near-full recovery produces exactly the realized vol that’s now running ahead of what the Nasdaq’s options currently charge. The S&P sits in between. It touched −3.42% against its own −3.51% line the same week the Nasdaq broke through, then spent the back half of the month working back to +2.41%.

None of this predicts anything. The bands are built from the same 20 sessions they’re measuring, so the path mostly staying inside them is closer to arithmetic than insight. What the picture adds is the shape behind the number: which lane ground there steadily, and which one took a single hard hit and spent three weeks working back from it. That’s worth knowing before sizing off HV20 alone, and it’s a different fact than the VRP percentile tells you.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike VIX they price this specific week, catalysts and all.

This week's ±1σ cones through Friday 8/14. Every cone narrowed again; CPI Wednesday and PPI Thursday both land with room to trade before settlement.

Every cone narrowed again, an eleventh to a fifth. That’s four straight Sundays of narrowing on at least two of the three lanes, and this week it’s all three.

The structural note this week is different from last week’s. Last Sunday’s cones settled the same session the jobs report printed, with almost no time to digest it. This week CPI lands Wednesday and PPI Thursday, both with at least a full session of cash trading behind them before Friday’s settlement. Whatever these prints do, the market has time to work through it before the cone closes — a different shape of risk than last week’s, not a smaller one.

What it means for premium sellers

Educational framing, not a trade — conditions, not signals.

• Record-low IV rank and record-low cushion are not the same claim, and this week they split. The S&P has both. The Russell has only the first — its options are the cheapest they’ve ever registered here, but the variance premium behind them is still 36th-percentile paid, not single digits. Cheap and thin aren’t interchangeable words.

• A negative variance premium is a statement about the last 20 sessions, not the next five. The Nasdaq has delivered more than it’s charged for four sessions running. That doesn’t forecast Wednesday; it means the standard pricing window is currently behind the tape, and worth checking again after CPI prints.

• This week’s two macro prints land with room to move, not on top of settlement. CPI Wednesday and PPI Thursday both leave a full session or more before Friday’s close. Compare that to last week, where the jobs report and the cone settlement shared a session with six and a half hours between them.

• A record read is a description, not a forecast. I don’t know if 8.1 or 1.8 holds through Wednesday — records like this tend to be interesting precisely because they don’t sit still for long. That’s an observation about how these numbers behave, not a call on which way they move.

• Check which window you’re sizing off before you lean on any single lane. HV20 says the Nasdaq’s premium is negative. That’s the standard window here, and it’s worth knowing you’re reading a 20-session average into a five-session catalyst.

The trap this week is the mirror of last week’s. Last Sunday the coast looked calm and the cushion was thin. This week the readings are the cheapest they’ve been all year, and there’s a genuine two-print inflation week sitting right behind them.

The risk lens

What a careful seller actually watches this week:

• CPI Wednesday, 8:30 AM ET. The week’s top-tier macro node, and the one most likely to move the vol complex on its own.

• PPI Thursday, 8:30 AM ET. Usually second-order to CPI, but a genuine two-day inflation read together with it — not the single print this dataset has mostly seen since late June.

• Retail Sales Friday, 8:30 AM ET. Prints the same session these cones settle; less market-moving than CPI historically, but it’s the last data point before the week closes.

• Cisco reports Wednesday after the close — the single-name print carrying the most index weight this week, and it lands the same day as CPI.

• The Nasdaq’s variance premium has been negative for four sessions. Worth re-checking after CPI, not before — a fifth negative session on the far side of an inflation print would be a different kind of data point than four heading into one.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Mon 8/10 — No top-tier macro release. A heavy day for smaller-cap earnings.

• Tue 8/11 — CoreWeave reports after the close.

• Wed 8/12 — CPI, 8:30 AM ET. Cisco reports after the close.

• Thu 8/13 — PPI, 8:30 AM ET. Weekly jobless claims the same morning.

• Fri 8/14 — Retail Sales, 8:30 AM ET. This week’s cones settle against Friday’s close.

The cheapest reading either the S&P or the Russell has posted all year, and a two-print inflation week to test it against. I don’t know which one is right about the week ahead. That’s what Wednesday’s Recalibration is for.

Your turn: The S&P’s IV rank and its variance premium hit record lows together this week; the Russell’s IV rank did too, but its cushion didn’t follow. Does that difference change how you’d size the two, or is a record-cheap read a record-cheap read regardless of what’s behind it? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

XOXO,

Theta Desk

Subscribe now

Welcome! Theta Desk is a data-driven publication about selling options. Expected-move levels, IV rank and regime reads three times a week, plus the deeper analysis of when selling premium actually works — and when it quietly stops.

I pull the numbers every session because I trade off them, and I publish what they say, including the weeks they say the strategy did worse than advertised.

If you only want one thing: the weekly levels are in the Weekly Outlook section. The research starts can be found how often the expected move actually holds. The mechanics, from zero, start with The Seller’s Side, Part 1.

Subscribe now

Why this exists

There’s one question underneath everything here: can selling options premium, done with discipline and honest accounting, become a real income engine, the kind you’d trust with a financial-independence plan — and under what conditions does it stop paying?

Most content in this corner of the market never gets near that question, because answering it honestly means publishing the parts where the strategy loses. Assignment drawdowns. Sequence risk. The weeks where the correct move was to sell nothing at all. Plenty of newsletters will show you winners. Almost nobody will quantify when premium selling stops paying.

So the standing rules here are boring and load-bearing: methodology gets shown, misses get published at full magnitude, and when the data disagrees with the strategy, the data wins.

What actually shows up in your inbox

Primarily three kinds of things, (with some exceptions).

1. The Weekly Regime Read — Sunday, Wednesday, Friday

This is the backbone, and the part you can set your week by. Three short, data-led reads of where volatility sits and what it’s paying:

• The Weekly Setup (Sunday) — the fullest read. Where the VIX complex is, which index has rich premium and which has a real cushion behind it, the expected-move cone for the week ahead on the S&P, Nasdaq and Russell, and exactly which catalysts land inside those cones.

• The Midweek Recalibration (Wednesday) — Sunday’s plan against what actually happened. What the market has already spent of its cone, what’s left, and how the setup adjusts from here.

• The Weekly Debrief (Friday) — the wrap. Did the cone hold, and if it did, was the ride survivable?

Those last two are not the same question, which is the sort of thing this publication exists to keep saying. In the week of July 24th, all three indexes closed inside the cone I’d drawn on Sunday. The Nasdaq traveled 8.39% to end down 1.6% — roughly 2.9 times its own weekly cone, in transit. The endpoint seller had a fine week. The seller who got margin-called on Wednesday did not.

That three-times-a-week read is free, and it’s what the subscribe button is for.

2. Options Data Studies — roughly every other week

These are the reason to stick around. Each one takes a rule everybody repeats, tests it against twenty years of SPX, and shows the code and the caveats. Three are live:

• How often does the expected move actually hold? — the textbook says 68%, and a band with no premium in it does exactly that. The VIX-priced band holds about 83%. That gap is the seller’s entire paycheck, and when the band breaks it breaks by about half a band.

• Which delta should you actually sell? — every rung from the 30 down to the 3. From the 25 out to the 12 it’s a plateau, not a slope: you’re choosing a position size, not an edge.

• Should you close winners at 50%? — held to expiry makes more money per trade at every DTE, which is the number everyone quotes. Score it per day of capital instead and the ranking flips, all five DTEs at once.

Every study adds rows to a regime dataset I’m building in the open — win rates and expectancy sorted by volatility environment. That dataset is the long game here.

3. The Seller’s Side — the ground-up track on options education

A slower series that builds the mechanics from zero, for anyone who’d rather not nod along to “vega” for another six months. It starts with what an option actually is and works up through pricing, the greeks, spreads and assignment.

Why the S&P, Nasdaq and Russell — and not single names

Everything here trades off three index products: SPX, NDX and RUT. That’s a deliberate scope, not a gap, and if you’re used to reading about earnings plays and single-stock spreads it’s worth saying why.

Index options settle in cash and carry no assignment risk on the underlying shares, no earnings gaps, no CEO tweet risk. The premium you collect is compensation for volatility, not company-specific news you’d need to track separately for every ticker in a portfolio. That makes the edge measurable — I can build twenty years of regime data on the S&P because its options have traded liquidly that whole time, priced off one thing: how much the market expects the index to move.

The three indices aren’t redundant with each other, either. The S&P is the benchmark — the deepest, most liquid options market there is, and the one every regime read is calibrated against. The Nasdaq runs hotter, carrying more tech concentration and a wider expected-move cone most weeks. The Russell is the small-cap read, thinner and more prone to a rich-IV, thin-cushion setup than the other two. Reading all three side by side is how you tell a broad volatility regime from a sector-specific one.

Single-name premium selling isn’t wrong, and the mechanics taught in The Seller’s Side apply just as well to a stock as an index. It’s a different game, though — one priced on earnings calendars and company idiosyncrasies rather than a clean volatility signal, and not the one this publication is built to measure.

Three words you’ll see constantly

• The expected-move cone — the range options pricing implies for the week, quoted from the actual straddle rather than scaled from an annual number. Prices, not predictions.

• IV rank — where today’s implied volatility sits inside its own past year. The number premium sellers anchor to, and here’s why.

• The cushion (variance risk premium) — how much more volatility was charged for than actually showed up. High IV rank with a thin cushion means you’re being paid a lot for a risk that’s genuinely there, and those two numbers disagree more often than you’d think.

Where the numbers come from

I pull the regime data myself every session and log it, so the record is forward-looking rather than reconstructed after the fact. For the option-level work I use real options pricing data rather than scraping estimates, then checked my reconstruction against Cboe’s own published settlement values.

What this isn’t

Theta Desk is education and analysis, every time. Never signals, never “buy this,” never a trade alert. I also don’t do technical analysis here. Nothing against it; it just isn’t what this data supports, and I’d rather stay inside what I can defend.

What’s left, when you take the signals and the P&L screenshots away, is the reasoning. The aim is a sharper and more independent trader — someone who can look at an IV rank of 69 with a thin cushion behind it and decide for themselves, rather than waiting for me or anyone else to say go.

If you’re new, read these four

1. How often does the expected move actually hold? — the foundation everything else sits on.

2. The most recent Market Outlook — to see how the market is positioned for options sellers.

3. The Seller’s Side, Part 1 — if any of the above lost you.

Subscribe and the Sunday setup lands before the week starts. If something here is wrong, say so in the comments — the correction runs in the open, same as the rest of the work.

Subscribe now

Theta Desk publishes educational market analysis only. Nothing here is financial, investment, or trading advice, or a recommendation to buy or sell any security or options contract. Options trading involves substantial risk of loss and is not suitable for all investors. Historical regime statistics and past performance do not guarantee future results. Do your own research and consult a licensed financial professional before trading. The author may hold positions in instruments discussed.

Subscribe now

TL;DR

Every Sunday I draw a one-standard-deviation range around each index for the coming week, straight off the straddle that expires Friday. Roughly two-in-three odds of finishing inside. I’ve been calling these cones since June, which is my shorthand and not the field’s — read it as “the range the options market priced.”

Sunday the S&P’s was ±1.31%, across five sessions.

Monday it moved 1.48%. Tuesday it moved 1.79%. Either one of those, on its own, was bigger than the entire week’s priced range. Then Wednesday and Thursday came in at −0.17% and −0.18%, which between them is less than a fifth of one percent, and Friday added 0.62% on the jobs report.

That’s the week. Two sessions did all of it and three sessions did nothing, and the two that did it were both big enough to break a five-day range by themselves.

All three indexes finished outside: S&P +2.72σ, Russell +1.92σ, Nasdaq +2.04σ. The first 0-for-3 since I started grading these. Tuesday broke the range on all three at once; the S&P had already broken its own on Monday’s close, session one of five.

And the S&P did it going up. It set record closes Tuesday and Friday — 7,757.64 on Friday is the highest it has ever closed — while finishing 2.72 standard deviations away from where the options market said it would. A rally to a new high is supposed to be the boring outcome. It wasn’t the boring outcome.

The week in vol

The gauges barely moved, which is its own kind of strange given what the tape was doing underneath them.

• VIX: 15.99 → 15.86 → 16.50 → 15.81 → 15.15 → 14.90. Net −1.09 on the week. The only up-day was Tuesday, the session the S&P rose 1.79% to a record close, which is a mildly funny place for the fear gauge to have its best day.

• The 9-day gauge did move. VIX9D went 13.05 → 13.28 → 15.04 Tuesday → 13.79 → 12.66 → 11.96. The Tuesday jump is the front end pricing the jobs report three sessions out; the 11.96 close is what’s left after the calendar emptied. That’s the lowest 9-day print in this scorecard’s short life.

• The curve steepened again. The 9-day-to-3-month gap went 5.97 → 6.76, with the back end essentially parked (VIX3M 19.02 → 18.72). Two Sundays ago that gap was 2.89. It has now widened for a third read running, entirely because the front keeps falling.

The front end priced the jobs report on Tuesday and gave it all back plus some. VIX9D closed the week at 11.96, the lowest print this scorecard has recorded.

Both index gauges did the same thing. VXN went 26.00 → 22.82, RVX 20.27 → 18.42, and neither had a single session that moved more than a point and a half. This was not a week the vol complex found interesting.

The scorecard: did the market stay inside the range?

The range settles on Friday’s close. That’s the only bell that counts for this table.

Zero for three. The S&P was already outside on Monday's close, with four sessions still to run.

The S&P closed Monday at 7,600.50 against a 7,587.8 ceiling. Twelve points and change past a line that was supposed to hold for five sessions, on day one, with four sessions still to run. By Tuesday’s close all three were out, the Russell by 51.98 points and the Nasdaq by 749.26, and none of them came back. Wednesday and Thursday were both mildly red on all three and it didn’t matter; the hole was already too deep to climb back into.

Here is the same fact from the other direction, and it’s the one I’d hang the week on. Every index had at least one single session that moved further than its whole week was priced to move.

Tinted where a single session moved further than the index was priced to move across all five. Tuesday did it on every index; the S&P did it twice.

Tuesday did it on all three. The S&P did it twice. And Monday’s 1.48% was 2.4× the S&P’s own daily expected move of ±0.62% — the straddle for that specific session, priced Sunday night.

Sunday’s issue spent its structural note on the fact that July payrolls printed Friday at 8:30 AM, on the same session these ranges settled, and said that was “the cleanest expiry decision the calendar has offered in weeks.” That was correct about the calendar and useless about the week. Payrolls contributed +0.62% on the S&P, +1.10% on the Russell, +1.19% on the Nasdaq — a normal Friday. The week was decided Monday and Tuesday, on no scheduled catalyst at all: the S&P and the Dow put in record closes on Tuesday as Treasury Secretary Bessent said a deal to reopen the Strait of Hormuz could land within a day or two, Brent fell 2.34%, and Palantir and Caterpillar carried the earnings tape. Confirm the macro items against your own calendar and your own sources — I’m relying on the wires for the Hormuz timeline.

For the record, the print itself was ugly: nonfarm payrolls fell 23,000 against a +83,000 consensus, with government down 53,000 and private payrolls up 30,000. Unemployment ticked to 4.1% on a shrinking labor force, and average hourly earnings slowed to 3.2% year over year, the softest since 2021. The market went up on it. I’d assume that’s rate-cut positioning, but I’m inferring, and one session is a thin place to read intent.

Wednesday’s Recalibration priced the return trip. It quoted the Friday straddle directly and found that getting back inside the range cost 1.84× the S&P’s entire remaining expected move, 1.19× the Russell’s and 1.35× the Nasdaq’s. When the repair job costs more than the whole move still priced, it usually doesn’t happen. It didn’t.

The running record, six Sundays priced and five Fridays graded:

Six Sundays priced, graded. 12 of 18 inside — 66.7%, which is now within a percentage point and a half of what a ±1σ range is built to do.

12 of 18 — 66.7%. A ±1σ range is designed to hold about 68% of the time. Last Friday the record read 80% and I wrote that the ranges had been “running a little generous.” One week later it sits within a percentage point and a half of the theoretical number, because a single 0-for-3 week is enormous when your denominator is 18. That’s the actual lesson of this table right now, and it’s about sample size rather than about the market. Ask me at fifty.

What the seller actually collected

What each lane charged Sunday against what the week actually delivered. Two of these five sessions each broke a five-day range on their own — and every range broke anyway.

Now put that table next to the one three sections up, where Monday and Tuesday each moved the S&P further than its whole week was priced to move. The standard measure says the S&P realized 14.35 against the 15.99 it charged. Two of those five sessions were violent enough to break a five-day range on their own, and the summary statistic for the week came out below what was charged for it.

I nearly wrote a cheerful paragraph about the variance seller having a fine week off that number before I went back and looked at the sessions. The number is not describing a calm week. It’s a statistic doing two things to a concentrated one.

First, it averages. Three of the five sessions moved −0.17%, −0.18% and +0.62%. Those near-zero days pull the estimate down hard, and the two that mattered get diluted by them.

Second, it subtracts the drift. Standard realized vol measures how much the daily returns scattered around their own average, and this week’s average return was strongly positive on all three indexes. So the trend — the exact thing that broke the ranges — gets removed from the number that’s supposed to describe the week. Take it out and you get the figures above. Leave it in and you get these:

The same five sessions, measured two ways. The gap between the columns is the trend — and on the S&P it flips the verdict from 1.11× to 0.94×.

That gap is the trend. On the S&P it flips the verdict outright: 1.11× becomes 0.94×, paid-fairly becomes underpaid, and the only thing that changed is whether the estimator is allowed to notice that the week went one direction.

Which of the two is right depends on what you sold. If you delta-hedged, the drift genuinely isn’t yours; you were trading the scatter, and the scatter was tame. If you sold a range and held it, the drift is the entire story and the left-hand column is describing somebody else’s week. Most retail premium selling is far closer to the second. I use the standard measure in this table every week because it’s comparable across weeks and it’s what the vol indexes are quoting against, and this is the first week where I think it’s actively misleading on its own.

One honest caveat on all of this. Five daily returns is a terrible volatility estimate, and I’m comparing them to 30-day implied gauges that aren’t pricing this week specifically. Read the direction, not the decimal. What I’d defend is the sign and the ordering; I would not defend 1.11× against 1.06× as a real difference.

The variance premiums themselves kept compressing. The Nasdaq’s is at 0.87×, the 2nd percentile of its own year — VXN at 22.82 against 26.10 of 20-day realized, charging less than the recent tape has delivered. The S&P is at 1.04×, the 6th percentile. Sunday’s issue printed 1.28× and 1.50× and called two of three lanes bottom-quartile. They’ve since gone lower.

The path: two sessions did it, and three sat there

Total distance covered versus net change. Last week these efficiencies were 14%, 0.6% and 4% — same measurement, same indexes, seven days apart.

Efficiency is net move divided by total distance traveled — how much of the week’s movement went somewhere. Last week’s readings were 14%, 0.6% and 4%. This week: 60%, 48%, 48%.

Same measurement, same three indexes, seven days apart, and roughly a tenfold difference. Last week the Russell traveled 7.51% to finish 1.34 points from where it started. This week it traveled 7.32%, essentially the identical distance, and finished 3.52% higher. The distance covered was the same. Where it ended up was not.

That’s the whole week in one comparison. Not more distance — the S&P actually covered less ground this week than last, 6.00% against 7.40%. What changed is that the distance arrived in two same-signed sessions instead of being spread across five that argued with each other. Concentrated and pointing one way is the shape that breaks a range, and it’s also the shape a five-day volatility average is worst at describing.

What it means for premium sellers

Education, not signals — conditions, not calls.

• Both straddles were run over; only the summary statistics stayed calm. The weekly range broke and the Monday daily range broke by 2.4×. What came in low was the 30-day gauges and a five-day average — measurements, not prices you could have traded. If your sizing came off “how volatile is this week going to be,” the answer you’d have gotten from VIX at 15.99 was not wrong so much as answering a different question than the one Monday asked.

• Watch the first session’s close against the whole week’s range. The S&P was outside on Monday. That doesn’t decide anything by itself (the Nasdaq closed below its range on Fed Wednesday last week and finished comfortably inside), but it’s the earliest honest read you get, and it’s free.

• A repair cost above 1.0× the remaining expected move is the market telling you something. Wednesday’s numbers were 1.84×, 1.19× and 1.35×, meaning the move required to get back inside was larger than the entire move still priced. None of the three made it. That’s one observation, not a rule, and I’ll keep quoting the figure to find out whether it’s worth anything.

• The premium is now the cheapest it has been all year. The Russell’s IV rank is 1.8, the 0.4th percentile of its own past year, after a week in which it traveled 7.32% and finished 3.52% higher. S&P IV rank 8.1, Nasdaq 36.2. That combination is a condition to size around, and reasonable people will read it in opposite directions.

• Six Fridays, six different lessons. A seller’s week, a buyer’s week, a week that punished over-adjusting, a week that split the endpoint from the variance, and now a week that arrived in two sessions and left the summary statistics looking innocent. I keep the table instead of the moral because I don’t yet know which of these repeats.

The week ahead

The full read is Sunday, so this is the shape only, and I’d rather price it off Sunday’s data than guess here.

Implied vol is lower across the board than it was seven days ago: VIX 15.99 → 14.90, VXN 26.00 → 22.82, RVX 20.27 → 18.42, and IV ranks of 8.1 / 1.8 / 36.2. Realized went the other way again — HV20 rose on all three for a second straight week (S&P 12.48 → 14.37, Russell 13.51 → 15.65, Nasdaq 23.65 → 26.10). That’s the same both-ends compression I flagged last Friday, now one notch further along, with the Nasdaq’s premium underwater.

The calendar ahead is lighter than the one behind. Whether the ranges that get priced against it are wide enough is Sunday’s job.

Your turn: Monday moved the S&P more than its whole week was priced to move, on nothing the calendar had flagged, and Wednesday and Thursday together moved it a third of a percent. If you sell weekly premium, does a session like Monday change what you do on Tuesday, or is a five-day range a five-day range and you sit in it? Reply — I read and answer every one, and the best reads make it into Sunday’s Setup.

XOXO,

Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday’s Setup drew a one-standard-deviation range on each index through Friday — the expected move, priced off the straddle, with roughly a 2-in-3 chance of containing the week. I’ve been calling that range the cone, and I’ll keep doing it here. Sunday’s issue spent most of its length on how thin the cushion behind it had gotten. Two sessions later the cushion isn’t thin. It’s behind us.

All three indexes closed Tuesday above the top of it. The S&P at 7,736.52 against a 7,391 – 7,588 range. The Russell at 3,036.98 against 2,878 – 2,985. The Nasdaq at 29,733.16 against 27,564 – 28,985. Measured in Sunday’s own units the S&P sits at +2.51σ — two and a half weekly expected moves, spent in two of five sessions.

What I didn’t expect was the shape of it. Last Wednesday the cones broke too, and they broke at 10–32% efficiency: the tape covered enormous ground and finished where it started. This week efficiency is 89% on the S&P, 91% on the Russell, 86% on the Nasdaq. Last week the market wasted its move; this week it spent it in one direction.

And vol went up through it. VIX 15.99 → 16.50, the 9-day gauge 13.05 → 15.04, on two sessions of record closes. Rallies don’t usually bid the front end. Friday morning is the first place I’d look for why.

Where we are vs. Sunday’s cones

Sunday’s ranges run through Friday’s close, roughly a 2-in-3 chance of finishing inside. Two of five sessions are gone. The band is Sunday’s cone, the solid bars are each session’s actual high-to-low range, the markers are the closes, and the shaded block on the right is what the market is currently charging for the three sessions that are left.

Sunday's ±1σ cones through Friday 8/7, with each session's actual high-to-low range drawn as a bar. All three closes are above the top. The shaded block is the Wed–Fri move quoted from the 8/7 straddle, not scaled — on all three, the distance back inside is larger than it.

The S&P was already out on Monday. It closed at +1.13σ on the first session of the week, which is outside a cone drawn to hold two-thirds of Fridays, and then added another 1.38σ on Tuesday. The Russell finished Monday at +0.94σ and the Nasdaq at +0.71σ, so both of those broke on session two.

Three sessions remain, and Friday 8/7 is a listed expiry on all three indexes, so I can read the price of exactly the stretch that’s left rather than inferring it: ±1.04% on the S&P, ±1.44% on the Russell, ±1.86% on the Nasdaq, Wednesday through Friday. That’s the denominator in the last column, and it’s what makes the number worth printing. Getting the S&P back inside Sunday’s range by Friday’s close takes a 1.92% decline against a 1.04% move still priced — 1.84 times the entire remaining budget, in one direction, in three days. The Russell needs 1.19× and the Nasdaq 1.35×.

None of that is a forecast, and it cuts the other way too: nothing says these indexes have to come back. The point is narrower. On Sunday the question was how much room was left inside the range. It isn’t a room question anymore. The market has spent three sessions’ worth of pricing getting somewhere the cone said it probably wouldn’t be, and the cone has three sessions to be wrong in.

One method note, because last week it mattered a lot and this week it barely does. Scaling Sunday’s weekly move by the square root of the sessions left would give 1.02% / 1.42% / 1.94% against the quoted 1.04% / 1.44% / 1.86%. Last Wednesday that same comparison was off by 13–15% on every index and flattered every cushion, because the FOMC, PCE and four mega-caps were stacked into the back half of the week. This week the calendar has one event on one morning, and the approximation nearly lands. Same method, different week, and the only way to know which kind of week you’re in is to quote the contract.

What actually did it

Monday the price of Brent fell 4.7% to $83.77 after the President said over the weekend he’d hold off on new strikes against Iran at the urging of regional allies, and the S&P added 110.78 points to 7,600.50 while the Dow put on 693. Tuesday was bigger — the S&P closed at a record 7,737, up 1.79%, the Dow up 1.71% to 54,086, on reports of progress toward reopening the Strait of Hormuz. Mega-cap tech carried both sessions: Meta +6%, Amazon +4% through a $3 trillion market cap, Alphabet and Microsoft close to +5% each, Nvidia +3%.

Sunday’s issue said the earnings slate had moved down-cap and there was less single-name gap risk inside the Nasdaq’s cone than in any week since mid-July. That was true and it didn’t matter. The move came from an oil headline and a geopolitical one, neither of which is on an economic calendar, and the Nasdaq covered 5.16% anyway. I’d like to pretend I have a lesson there beyond “the calendar only lists the risks somebody scheduled,” but I don’t.

The distance, and how straight it was

Week-to-date efficiency, this Wednesday against last. Efficiency is net move divided by total distance traveled. Both weeks broke Sunday's cone by the second session — at 10–32% last week and 86–91% this week. Two elapsed sessions in each window.

Efficiency is net move divided by total distance covered — how much of the walking turned into progress. At 89% the S&P gave back almost nothing on the way. There were no round trips to sell into and no mean reversion to wait for, which is the version of a cone break that a short-premium book has the least defense against. A chopping tape at least hands you the same strike twice.

Two sessions is a thin window and I’d hold the decimals loosely. But 86–91% across three separate indexes isn’t a rounding artifact, and it is a genuinely different regime from the one I was writing about seven days ago at 10–32%. Both weeks broke the cone. I don’t know which one is more common after a vol crush, and that’s a real gap: with five Sundays of cone history I can tell you how often the endpoint lands inside, and almost nothing about the path that got there.

What changed in vol

Same-method comparison — each gauge against Sunday’s read, nothing rebased.

The nine-day gauge did four times the work of the thirty-day one and six times the three-month. Sunday’s curve had re-steepened to 5.97 points front-to-back, the widest in over a month; two sessions of rally flattened it back to 4.30, and the flattening came entirely from the front.

Vol rising into a melt-up is the part I’d flag. Monday behaved normally — VIX slipped to 15.86 on a 1.5% rally, which is the usual arrangement. Tuesday the index gained 1.79% to a record close and VIX went to 16.50 with VIX9D up 1.76 on the session alone. The front end is where Friday’s 8:30 print lives, and a nine-day gauge repricing hardest is the cleanest read available for what the options market thinks is actually in front of it. That’s the read I’d give it. I’d also note that two sessions of unusually large up-moves mechanically feed the same realized-vol number the front end tracks, so some of this is arithmetic rather than opinion, and from here I can’t cleanly separate the two.

The cushion Sunday flagged, and where it went

Sunday’s lane table printed the Nasdaq’s variance premium at 1.10× on HV20, then flagged that on HV10 it was already 0.97× — implied below its own ten-day realized. I said I wasn’t going to swap windows mid-issue to find a scarier number.

Two sessions later the slower window agreed on its own.

VRP percentile, Sunday to Tuesday. All three fell, all three for the same reason: implied barely moved while 20-day realized rose on every index. The Nasdaq crossed below 1.00× — it now charges 25.48 against 26.19 delivered.

The Nasdaq now charges 25.48 vol against 26.19 delivered over twenty sessions. That’s a negative variance premium on the standard window, the 10th percentile of its own year, and it got there from both ends at once: VXN fell 0.52 while HV20 rose 2.54. On the ten-day window it’s 0.81×.

The S&P slid 25th percentile to 16th and the Russell 80th to 52nd, and the mechanism is the same everywhere: realized vol is climbing faster than implied. Every HV20 on the board rose, and every HV10 rose more. Two enormous up-sessions are still vol — the realized calculation doesn’t care which way you moved, and a 1.79% day feeds the same number a 1.79% drop would.

Which leaves the Russell as the only lane where the numbers still agree with each other. IV rank 13.6, the 9.5th percentile of its own year, and a 1.28× cushion at the 52nd. Cheapest headline premium on the board, median cushion behind it. That’s a smaller edge than the 80th percentile it carried on Sunday, and it’s the only one of the three that didn’t get worse than ordinary.

The adjustment

• A cone can be broken on Wednesday and inside on Friday. These settle against Friday’s close, not against Tuesday’s, and the running record here is 12 of 15 inside across five Sundays. Nothing about a mid-week breach is a scorecard entry yet. It does mean the position that has to be reversed is now larger than what’s priced.

• The direction of the break doesn’t change the arithmetic. A short strangle marked against +2.51σ is marked against +2.51σ. Nothing here is a comment on whether stocks should be higher; the cone was a range and the tape left it.

• High efficiency is the harder version to hold. Last week’s break came with 10–32% efficiency, which bleeds a book through adjustments while the closes stay boring. This one has 86–91%, which is a one-way move against whichever side was short it. Different problem, and the second one arrives faster.

• The Nasdaq’s premium is negative on the standard window. 25.48 charged against 26.19 delivered. If you size off the variance premium on NDX, go check what number you’re actually sizing off this week.

• Map the expiry to Friday 8:30. Anything living past Thursday’s close is short the July employment report with one session to resolve it, and the front-end vol move says the options market is treating that as the week’s real content. Anything expiring Wednesday or Thursday isn’t.

I don’t know whether the front end is pricing payrolls or just catching up to two big realized sessions. Both stories fit Tuesday’s tape exactly, and they stop fitting equally well at 8:31 Friday morning.

What’s left on the week’s calendar

• Wed 8/5 — ISM Services PMI, 10:00 AM ET. Eli Lilly and Disney report before the open. One calendar I checked listed ISM Services on Tuesday rather than Wednesday, so this is worth a look at your own before you lean on it.

• Thu 8/6 — Weekly jobless claims, 8:30 AM ET.

• Fri 8/7 — July Employment Situation, 8:30 AM ET. Consensus runs around 120k nonfarm payrolls against June’s 57k, with the unemployment rate expected to tick to 4.3% from 4.2% and average hourly earnings +0.3%. Those are forecasts collected from one preview and worth verifying against your own source. Sunday’s cones settle against Friday’s close, so the print and the settlement share a session.

Friday’s Debrief writes the scorecard, including whether three cones that broke on the way up on Tuesday were still broken when it counted.

Your turn: All three indexes are above the top of Sunday’s ±1σ cone by Tuesday, at 86–91% efficiency, after last week broke the same cones at 10–32%. Which of those two weeks is harder on a short-premium book: the one that goes nowhere loudly, or the one that goes one direction fast? Reply — I read and answer every one, and the best reads make Friday’s Debrief.

— Theta Desk

Subscribe now

Subscribe now

TL;DR

Last week’s event calendar is cleared. The Fed held, core PCE printed, Apple and Amazon got through the door, and vol did what it does after all of that: VIX 18.58 → 15.99, the 9-day gauge down from 20.38 on Thursday to 13.05. Every weekly cone on the board narrowed. If you sell premium, the tape looks calmer and the credits look smaller, which is the normal trade-off and nobody would write an issue about it.

What makes this week different: implied vol fell on all three indexes and realized vol went up on all three. The S&P’s 20-day realized went 10.41 to 12.48. So the variance premium, which is what you’re paid measured against what the index is actually delivering, got squeezed from both ends at once. The S&P’s was the fattest cushion on the board a week ago at 1.78×, the 79th percentile of its own year. It’s now 1.28×, the 25th. The Nasdaq sits at 1.10× and the 24th.

Cheaper premium is fine. Cheaper premium against faster movement is a different thing, and it’s what’s on the board. The one thing to internalize: two of the three lanes now pay you a bottom-quartile cushion, and the week’s only real macro node, Friday’s jobs report, prints on the same session these cones settle.

The regime right now

Last Sunday I wrote that every catalyst was parked inside the Nasdaq’s cone. They came and went, and the vol complex emptied out behind them.

• VIX 15.99 — down 2.59 from last Sunday’s 18.58, and down 4.67 from Thursday’s 20.66 intraweek high. Back in the lower half of normal.

• Term structure: 13.05 → 15.99 → 19.02 (9-day → 30-day → 3-month). Still contango. But the 9-day-to-3-month gap is now 5.97 points, up from 2.89 last Sunday — the steepest reading in over a month, and it re-steepened almost entirely from the front. VIX9D dropped 4.57 points; VIX3M dropped 1.49.

The curve, last Sunday vs. now. Everything fell; the front fell 4.57 points against the back's 1.49, re-steepening the 9-day-to-3-month gap 2.89 → 5.97.

Four Sundays running, the front end has been the part of the curve doing the talking. It rose into the Fed, spiked to 20.38 on Wednesday’s hold, and has now given back everything and then some. The back end barely participated in any of it, which is roughly what a well-behaved curve should do when the thing that moved was a scheduled event with a known date.

A steep front-to-back gap like this is usually read as calm: nothing priced near-term, more priced later. That reading is fine as far as it goes. It’s also a statement about what the options market can see on a calendar, and this week’s calendar is unusually empty until Friday morning.

The receipt from last week, since we keep those: all three closed inside their cones, and the ride was awful. The Russell traveled 7.51% of ground to finish +0.05%, six-tenths of one percent efficiency, which is about as close to running in place as an index gets. The Nasdaq covered 13.04% to net +0.52%. That’s the movement now sitting in the realized numbers below.

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

Every cushion on the board compressed in five sessions. The S&P lost 54 percentile points; two of the three lanes now sit in the bottom quartile of their own year.

Three things to read off this.

The compression is the story, not the level. Every VRP percentile on the board fell, and the S&P’s fell off a cliff: 79 to 25 in five sessions. That happens when implied drops and realized climbs simultaneously, which is exactly what a post-event week looks like. The vol index mean-reverts fast because it’s a forward quote and the event is behind it. The realized number can’t mean-revert fast, because it’s a 20-day window and last week’s chop is now stuck inside it for the next month.

The Russell is the best-cushioned lane and has the lowest rank on the board. IV rank 13.0, the 7.9th percentile of its own year — RVX is nearly as cheap as it has ever been. And its variance premium is 1.50×, the 80th percentile. Small-cap implied is low in absolute terms and still comfortably above what the index has been delivering. Worth flagging that this is the first Sunday we can say any of that with a straight face: RVX was never delisted, only our data lookup was broken, and the fix landed 7/26. Two Sundays ago this row said “HV30 stand-in, no VRP available.” It was wrong, and it’s fixed.

The Nasdaq’s number is still the biggest and still the thinnest. Rank 55.1, 26.00 vol, and a 1.10× premium in the 24th percentile. It charges 2.35 points more than the index delivered over 20 days. That’s the whole margin.

Now the honest part, and it complicates everything above. HV20 looks backward across twenty sessions, and the last ten have been faster than the twenty. S&P HV10 is 14.85 against HV20 12.48. Russell 16.05 against 13.51. Nasdaq 26.67 against 23.65, which means Nasdaq implied vol, at 26.00, is currently below its own 10-day realized. Measured that way the Nasdaq’s premium isn’t thin, it’s negative: 0.97×.

I’m not going to swap the metric mid-issue to make a point. HV20 is the standard here, it’s what every prior issue used, and switching windows to find a scarier number is how you talk yourself into a conclusion. But I’d be hiding something if I didn’t print the 10-day column, because all three point the same direction and that direction is realized vol is accelerating, not decaying. Whether that continues into a quiet week, I don’t know. It’s the number I’ll be checking Wednesday.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike VIX they price this specific week, catalysts and all.

This week's ±1σ cones through Friday 8/7. Every cone narrowed a fifth to a quarter, and the week's one macro node lands on the settlement session itself.

Every cone narrowed, by a fifth to a quarter. Same five-session horizon, much smaller invoice. Last week the straddle was pricing the Fed, PCE and four mega-caps into five sessions; this week it’s pricing ISM, a mid-cap earnings slate, and payrolls.

One structural note on those cones. Every one of them settles Friday, and the July employment report prints Friday at 8:30 AM. So the single largest scheduled item on this week’s calendar lands inside the cone with about six and a half hours of cash session left to resolve it. Compare that to last week, where the Fed hit Wednesday afternoon and the index had two full sessions to digest it, and used them, running the Nasdaq up 3.98% off Wednesday’s low.

None of that predicts Friday. It’s just where the risk sits in the week’s shape, and an event that arrives with the bell leaves you fewer sessions to be right in.

What it means for premium sellers

Educational framing, not a trade — conditions, not signals.

• A narrower cone and a thinner cushion are two separate facts, and both are true. The S&P’s weekly range shrank 26% while its variance premium fell 54 percentile points. You’re being paid less in absolute terms and less relative to delivered movement. Neither of those is a reason to do anything in particular; both are inputs to whether the credit is worth the risk this week.

• The Russell is the one lane where the two measurements agree. Lowest rank, best cushion, 1.50× at the 80th percentile of its own year. That combination is rare enough to be worth naming. It also comes with the widest cone of the three relative to its own history and the thinnest liquidity, and small-caps carry the payroll print harder than the mega-cap indexes do.

• Post-event vol crush is the most reliably mistimed thing on the calendar. The instinct after a cleared event is that the risk has passed and the premium is safe. What actually happened is the price of insurance fell while the thing it insures against sped up. That gap closes one of two ways.

• Map your expiry to Friday 8:30 AM. Anything living past Thursday’s close is short the jobs report with no session left to recover. Anything expiring Wednesday or Thursday isn’t. This is the cleanest expiry decision the calendar has offered in weeks.

• Check HV10 against HV20 before you size off the variance premium. On all three indexes the recent window is running hotter than the standard one. That doesn’t invalidate the VRP; it means the cushion you’re reading may already be stale.

The trap this week is straightforward: five quiet sessions on a calendar, cheaper options, and a natural conclusion that the coast is clear. The cushion numbers say the coast got thinner while the calendar got quieter.

The risk lens

What a careful seller actually watches this week:

• Friday 8:30 AM ET is the week. The July employment report is the only top-tier macro release on the schedule, and it prints on the session the weekly cones settle. Everything else is second-order by comparison.

• Two ISM prints frame the week. Manufacturing Monday at 10 AM, Services Wednesday at 10 AM. Neither typically moves an index the way payrolls does; both feed the rate path that payrolls will confirm or break.

• The earnings slate moved down-cap. AMD Tuesday after the close is the single-name print with the most index weight behind it, and it isn’t in the same category as last week’s four. For NDX specifically, that means less single-name gap risk inside the cone than any week since mid-July.

• Realized vol is running above its 20-day average on every index. HV10 exceeds HV20 across the board. If you’re sizing off the variance premium, you’re sizing off the slower of the two windows.

• A 25th-percentile cushion is a sizing input, not a verdict. Plenty of weeks with bottom-quartile variance premiums pay perfectly well, because the premium is the compensation, not the outcome. It just means less room for the week to surprise you before the math turns.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Mon 8/3 — ISM Manufacturing PMI, 10:00 AM ET. Palantir reports after the close.

• Tue 8/4 — Caterpillar before the open, AMD after the close.

• Wed 8/5 — ISM Services PMI, 10:00 AM ET. Eli Lilly and Disney report before the open.

• Thu 8/6 — Weekly jobless claims, 8:30 AM. Q2 productivity and unit labor costs the same morning.

• Fri 8/7 — July Employment Situation, 8:30 AM ET. This week’s cones settle against Friday’s close, so the print and the settlement share a session.

The credit got smaller and the cushion got smaller faster. Nothing on this week’s calendar looks dangerous, and the margin for error behind it is thinner than the calm tape suggests.

Your turn: The S&P’s variance premium went from the 79th percentile of its year to the 25th in five sessions, while its weekly cone narrowed 26%. Does a bottom-quartile cushion change how you size, or is the premium the premium and you take the week as it comes? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

— Theta Desk

Subscribe now

Subscribe now

TL;DR

Sunday I drew ±1σ cones around five sessions that had the FOMC, June PCE, and four of the five largest companies in the world parked inside them, and asked one question: is the Nasdaq’s cushion wide enough to hold that?

The week answered twice, and the two answers don’t agree.

On Wednesday the Nasdaq closed at 27,192.31 — below the 27,242 floor of its own cone. Not a wick, a close. The Fed held at 3.50–3.75% on a 9–3 vote with three governors dissenting for a hike, VIX jumped to 20.66, and the regime label flipped to Elevated for exactly one session. Then June core PCE came in at 3.3%, Apple and Amazon cleared, and the Nasdaq ran 3.98% off that low close in two sessions to finish at 28,274.20, +0.17σ.

So the scorecard reads 3 for 3 inside. S&P +0.59σ, Nasdaq +0.17σ, Russell +0.02σ — the Russell finished 1.34 points from where it started, on the week the Fed had three dissents.

And the vol seller still lost on the Nasdaq. It realized 32.37 annualized against the 28.39 it was charging Sunday — 0.88×, less than what it collected. The cone graded a hold and the variance sale didn’t, because those measure two different things, and this is the cleanest week I’ve had to show the difference.

The week in vol

One session did nearly all of it.

• VIX: 18.58 → 18.67 → 18.21 → 20.66 → 17.09 → 15.99. Flat into Fed day, a 2.45-point spike on it, then a 4.67-point crush across the two sessions after. Net −2.59 on the week, and Friday’s 15.99 is the lowest close in this scorecard’s short life.

• The front end did it harder. VIX9D went 17.62 → 20.38 Wednesday → 13.05 Friday. That’s a 7.33-point fall in two sessions. The 9-day gauge is the one that has to live through the events, and once the events were behind it, there was nothing left to price.

• The curve re-steepened. The 9-day-to-3-month gap had flattened three Sundays running — 7.42, then 3.69, then 2.89. It closed this week at 5.97. Front-end risk got priced, delivered, and then left the building; the back end barely moved (VIX3M 20.51 → 19.02).

  

  Fed day priced the risk and the next two sessions took it all back. The 9-day gauge fell 7.33 points from Wednesday's close to Friday's.

  Wednesday’s Recalibration went out a few hours before the decision, flagging that fed funds futures were pricing a live ~36% chance of a hike and that vol had fallen into it. The hike didn’t come. Three dissents did, which is close enough that the market repriced anyway — and then unpriced it completely once PCE cooled the next morning. I don’t think the front end was wrong to be nervous on Wednesday. I think it was wrong on Tuesday, when it drifted lower with all of it still ahead.

  ---

  The scorecard: did the market stay inside the cone?

  Every index carried a weekly expected move on Sunday — the market’s own one-standard-deviation estimate, roughly a 2-in-3 chance of finishing inside the range. The cone settles on Friday’s close, which is the only bell that counts for this table.

  

  The Nasdaq's Wednesday close was 49.89 points below its cone floor. The cone grades Friday, so it still reads Inside.

  

  Each bar is one session's actual high-to-low range inside the week's priced cone. The Nasdaq's Wednesday bar is the only one that breaks a rim — and it closed there, at −1.06σ.

  The Nasdaq’s Wednesday close sat 49.89 points below its cone floor — 0.18% outside, which is nothing as a percentage and everything as a fact. It traded to 27,176.03 intraday, −1.08σ. If you had been marked Wednesday afternoon on a short put structure at the lower rim, you were marked outside the range the whole week was drawn around, with two sessions left and Apple and Amazon still to report.

  Then Thursday happened. The Nasdaq gained 3.31% on PCE day, another 0.60% Friday, and the same cone that had been breached graded a comfortable hold. Nothing about the second half of that week was available to you on Wednesday at 4 PM.

  The running record, since this is the fifth Sunday I’ve priced cones and the fourth Friday I’ve graded them:

12 of 15, or 80%. A ±1σ cone is built to hold about 68% of the time, so five weeks in, the cones have been running a little generous — the market has been charging for slightly more movement than it’s delivered at the endpoint. That is the variance premium showing up exactly where you’d expect it. It’s also fifteen observations, which is not a sample you should build anything on. Ask me at fifty.

What the seller actually collected

Here’s where the week stops being simple.

The cone is a bet on where the index stops. A short options position is a bet on how much it moves along the way. Most weeks those two point the same direction and nobody has to think about it. This week they split, and the Nasdaq is the cleanest example I’ve had to write up.

Every variance premium on the board compressed, from both ends at once — implied fell and realized rose.

Implied = the 30-day forward gauge as of the 7/24 close. Realized = annualized close-to-close vol over the five sessions of the week itself. VRP is the standing 20-day read, Sunday against Friday.

What each lane charged Sunday against what the week actually delivered. The Nasdaq is the bar that inverts: 32.37 realized against 28.39 charged.

The Nasdaq realized more volatility than it charged for. 32.37 delivered against 28.39 priced. A seller who held Nasdaq vol through this week collected a premium that turned out to be too small for the movement they absorbed — and the index still finished 0.17σ from where it opened. Both of those sentences are true at once, which is the whole reason the travel column exists.

The S&P came out at 1.01×, which is a rounding error away from being paid nothing for the week. Sunday’s issue called it the cushioned lane at 1.78× on 20-day realized, the 79th percentile of its own year. I also wrote, in that same section, that HV20 looks backward and a fat-looking cushion “can be nothing at all” if realized vol stops falling. It stopped falling. S&P realized went 10.41 to 12.48 over the month and 18.38 on the week itself.

The Russell was the one that worked: 22.33 charged, 18.00 delivered, and it finished the week 1.34 points from its Sunday spot. Which is worth saying plainly — Sunday’s lane table printed a dash in the Russell’s variance-premium column and told you it couldn’t be measured. That was wrong, Wednesday’s issue corrected it, and the lane I said I couldn’t score was carrying a 2.02× premium at the 99th percentile of its own year going into the week. It was the best-cushioned lane on the board and I had it as unmeasurable.

One honest caveat on this whole table. I’m comparing a 30-day implied gauge to five daily returns. Five returns is a terrible volatility estimate; one 3.31% Thursday moves it several points on its own. And VIX, VXN and RVX are pricing a month, not this week. So read the direction, not the decimal. The 0.88× is directionally solid because it takes a lot of noise to flip a sign that far; the 1.01× on the S&P I would not defend to two places.

The path: what the adjustments paid for

Efficiency = net move ÷ distance traveled, across the five sessions of the closed week. Travel sums each session's true range, gaps included.

Total distance covered versus net change, the closed week. The cone prices the endpoint; travel is what your adjustments paid for.

The Nasdaq covered 13.04% of ground — 4.1 times the width of the cone it finished inside — to net half a percent. Last week it was 8.39% and I called that a lot. The Russell is the one that reads like a typo: 7.51% traveled, 0.05% net, 0.6% efficiency. It went out and came back so precisely that it landed within a point and a third of its starting price.

Wednesday’s Recalibration had already flagged this at the two-session mark, when all three had traveled more than a full cone’s width with every catalyst still ahead. The back half didn’t calm down. It doubled the distance.

What it means for premium sellers

Education, not signals — conditions, not calls.

• “Inside the cone” and “the vol sale worked” are separate results, and this week they came apart. The Nasdaq graded Inside and realized 0.88× what it charged. If your only weekly scorecard is the cone, you booked a win on a position that was underpaid for the movement it carried.

• A mid-week breach isn’t a loss on the scorecard, and it is absolutely a margin event. The Nasdaq spent Wednesday’s close outside a range it finished inside. Whether that mattered depended entirely on your size, not your thesis — the thesis was right on Friday and unhelpful on Wednesday.

• A backward-looking cushion is a description, not a promise. Sunday’s 1.78× on the S&P was computed against a quiet month. The week that followed wasn’t quiet, and the cushion measured against it was 1.01×. HV20 tells you what has been happening. It has no opinion about the Fed.

• The crush arrives after the event, not before it. VIX9D went 20.38 to 13.05 in two sessions once the calendar emptied. That’s a real and repeatable feature of event weeks, and it’s also why the credit you’re offered Sunday and the credit that’s there Friday are different numbers for reasons that have nothing to do with being right.

• One week, one sample. Ed. 1 was a seller’s week, Ed. 2 was a buyer’s week, Ed. 3 punished the over-adjuster, and this one split the endpoint from the variance. Four Fridays, four different lessons, which is roughly what I’d expect and exactly why I keep the table instead of the moral.

The week ahead

The full read is Sunday, so this is the shape only.

Every premium gauge on the board is materially cheaper than it was seven days ago. IV rank: S&P 29.1 → 14.3, Nasdaq 69.4 → 55.1, Russell 25.5 → 13.0. The S&P’s is the 23rd percentile of its own year and the Russell’s is the 8th. Realized vol went the other way on all three, so every variance premium compressed from both ends at once — the S&P’s from the 79th percentile of its year to the 25th in five sessions.

The calendar behind us was the densest of the quarter and the calendar ahead of us is not. What that combination means for the week’s cones is Sunday’s job, and I’d rather price it off Sunday’s data than guess at it here.

Your turn: The Nasdaq closed 49.89 points below its cone on Wednesday and inside it on Friday, and realized 32.37 vol against the 28.39 it charged. Do you grade that week as a win because the cone held, or as a loss because the variance sale was underpaid? I genuinely go back and forth on it. Reply — I read and answer every one, and the best reads make it into Sunday’s Setup.

— Theta Desk

Subscribe now

What this series is

Theta Desk publishes three data-driven issues a week (a Sunday setup, a midweek check, a Friday debrief), plus the occasional study where I take a piece of common wisdom and test it against twenty years of data. All of that assumes you already speak the language. It quotes IV rank at you. It discusses the expected move like you know what a straddle costs.

So there’s a third kind of post here now, and this is it. Twelve parts, from zero, one idea each, no prior knowledge assumed. The market reads are about this week and they’re finished by Friday. The studies test one specific claim and report what the data says. These twelve teach the machinery, and nothing in them depends on what volatility did last week — read Part 4 a month late and it’s still correct.

The goal is narrow, and I’d rather state it plainly than oversell it: by Part 12 you’ll open any issue this desk publishes and read it the way I write it, including the parts I don’t stop to explain.

Where it goes:

• Parts 1–3 — the contracts, and why anyone sells them. What an option is, where its price actually comes from, and the real economic reason sellers get paid to take the other side. Part 3 is the foundation for everything after it.

• Parts 4–6 — the greeks, from the short side. Theta is the paycheck, delta is the risk you’re paid to carry, vega is the thing that hurts you. Most courses teach these from the buyer’s seat. I won’t.

• Parts 7–8 — the two numbers the weekly issues lead with. IV rank and the expected move, fully decoded. After these, the Sunday setup stops being a wall of figures.

• Parts 9–11 — the structures, and the sizing. Cash-secured puts and covered calls first, then defined-risk spreads on indexes, which is where this desk actually spends its attention; Part 10 makes the case for why. Then sizing, which is the least interesting part of the series and the one that decides whether you’re still doing this in two years.

• Part 12 — a real Weekly Setup, annotated line by line, every callout pointing back to the part that taught it.

Each part also ends by showing you where the idea lives in an actual issue, so the series doubles as a tour of what the desk publishes.

One practical note: this isn’t a separate publication or a signup of its own. The parts arrive in the same inbox as everything else, grouped in their own section so the twelve stay together and stay easy to find later. What’s different is the kind of writing, not where it comes from.

Whether twelve is the right number I don’t know yet. If a part lands badly or leaves an obvious hole, I’d rather add a thirteenth than pretend the outline was right the first time.

The two contracts

Here’s the whole product in one sentence: you pay $2 a share today for the right — not the obligation — to buy a stock at $105 any time in the next 30 days, no matter where it trades between now and then. The stock is at $100. If it never reaches $105, you lose the $2 and that’s the end of it. If it runs to $115, your right to buy at $105 is worth $10, and your $2 became $8.

That’s an option. Everything else in this market — the greeks, the volatility surface, every strategy with a nautical-sounding name — is arithmetic built on that sentence.

There are exactly two contracts:

• A call is the right to buy something at a fixed price, on or before a fixed date.

• A put is the right to sell something at a fixed price, on or before a fixed date.

That’s the complete menu. Covered calls, credit spreads, iron condors, the wheel — every structure you’ll ever hear named is some combination of these two contracts, bought or sold.

The word doing the real work in both definitions is on the other side of the trade. The buyer of an option holds a right, and rights are optional — hence the name. Walk away, let it expire, nothing further happens. The seller of an option holds an obligation. Sell that $105 call and if the buyer wants stock at $105, you deliver stock at $105. There’s no walking away; you were paid the $2 precisely for giving that up.

Buyers pay for the right to change their mind. Sellers get paid for promising not to.

It’s an insurance market

The put makes this easiest to see. Say you own the stock at $100 and buy the $95 put. You’ve bought a policy: a $5 deductible (the distance to the strike), a $2 premium, a 30-day term. Whatever happens to the stock, you can sell at $95 until the policy expires.

Whoever sold you that put is running the insurance desk. They collect premiums, most policies expire unused, and now and then they pay out a claim. Theta Desk writes for the people running the desk — that’s the newsletter’s whole angle, and this series is named for it.

The analogy has a known crack in it: nobody buys fire insurance hoping the house burns down, while plenty of call buyers are simply making a fast, leveraged bet. Fine. Follow the money and the frame still holds — premiums flow one way on schedule, and claims occasionally flow the other, large.

The fine print, all of which matters

Four terms define any option contract:

• Strike — the fixed price the right applies at. $105 in our call example.

• Expiration — the date the right dies. After it, the contract is worth exactly its exercise value or nothing.

• Premium — what the buyer pays the seller for the contract, quoted per share.

• The multiplier — one standard equity option covers 100 shares. A “$2” option costs $200. A promise to buy stock at $95 is a $9,500 promise.

The multiplier is the one every new trader learns exactly once.

What the two payoffs look like

Read the left panel with me. The flat segment is the hole: below the $105 strike the call expires worthless and the buyer is out the $2 premium — no more, no matter how wrong the idea was. That flat floor is why buying options feels safe. At $105 the line kinks; at $107 (strike plus premium) it crosses zero, and past that the buyer is genuinely up.

Now the dashed line, because it’s the reason this series exists. The seller of the same contract holds the exact mirror image: their best case is the buyer’s flat segment — keep the $2, done. Their line falls as the stock runs. One side’s loss is the other side’s gain, dollar for dollar, at expiration. Hold that mirror in your head permanently: whenever this desk discusses a trade, the dashed line is where we’re standing.

What happens at the end

Three exits from any option position:

1. Sell it back (or buy it back, if you sold it) before expiration. This is what usually happens — options are traded far more than they’re exercised.

2. Exercise — the buyer uses the right; stock changes hands at the strike.

3. Expire worthless — out-of-the-money at expiration, the contract simply dies, and the seller’s job ends with the premium fully kept.

You’ll hear a statistic about how often options expire worthless. It gets quoted dishonestly in both directions — sellers wave it around as free money, skeptics dismiss it as survivorship. I’ll do that number honestly in Part 3, because it’s the foundation of the entire selling business.

Where you’ll see this in Theta Desk

From a recent Weekly Setup, discussing a week loaded with economic releases:

Selling the weekly here means carrying CPI, PPI, and the bank slate for a quiet-week credit.

Every word of that is Part 1 vocabulary. “The weekly” — options expiring that Friday. “A credit” — the premium, received rather than paid, because we’re on the selling side. “Carrying” — holding the obligation, the dashed line, through everything the week has scheduled.

Your turn: if you’ve traded options before, what’s the thing nobody explained to you until it cost money? Reply — the answers decide where this series slows down and adds a part.

Next in the series — Part 2: where an option’s price actually comes from, and why a contract that can’t be exercised profitably today still costs real money.

Theta Desk publishes educational market analysis only. Nothing here is financial, investment, or trading advice, or a recommendation to buy or sell any security or options contract. Options trading involves substantial risk of loss and is not suitable for all investors. Historical regime statistics and past performance do not guarantee future results. Do your own research and consult a licensed financial professional before trading. The author may hold positions in instruments discussed.

Subscribe now

TL;DR

Sunday’s Setup drew a one-standard-deviation range for each index through Friday and pointed out that every catalyst on the week’s calendar (the Fed, PCE, four mega-caps) was parked inside those ranges.

Two sessions later, none of those events has happened, and the tape has already covered more ground than the cones were drawn for. The S&P has traveled 2.25% against a ±1.78% weekly cone. The Russell 2.55% against ±2.33%. The Nasdaq 4.48% against ±3.15%. All three have spent more than a full week’s worth of distance in two sessions, and all three sit close to where they started: net +0.23%, +0.81%, and −1.30%.

So nothing has moved and everything has moved, which is the least comfortable version of a quiet tape. The cone prices where you finish; travel is what you pay to get there. This week the market has run through the travel budget before the events that were supposed to spend it.

The other thing that’s happened: vol went down. VIX 18.58 → 18.21, and VIX3M fell twice as hard. Fed funds futures currently put roughly a 36% chance on a 25bp hike this afternoon. I don’t have a clean story for why the vol complex is easing into a decision the rates market thinks is genuinely two-sided, and I’d be careful with anyone who does.

Where we are vs. Sunday’s cones

Sunday’s ranges run through Friday’s close — roughly a 2-in-3 chance of finishing inside. Two of five sessions are gone. The band is Sunday’s cone, the solid bars are each session’s actual high-to-low range, and the markers are the closes.

Sunday's ±1σ cones through Friday 7/31, with each session's actual high-to-low range drawn as a bar. Every close is inside; the Nasdaq's Tuesday low reached −0.76σ. The ranges are what the travel figure sums — a closes-only chart hides all of it.

Everything is inside, and for once they aren’t all leaning the same way — the S&P and Russell drifted up, the Nasdaq walked down.

Those are closes, and the closes undersell it. The Nasdaq finished Tuesday at −0.41σ but traded down to −0.76σ during the session, three-quarters of the way to the floor of a cone that still has three days to run. The S&P’s low tick was −0.22σ, the Russell’s −0.10σ. Nothing about that is a breach and none of it settles anything — the cone pays off on Friday’s close, not on the lows along the way. But if you were short Nasdaq puts on Tuesday afternoon you were marked against −0.76σ, not against −0.41σ, and those are different afternoons.

Three sessions remain, and they have their own price. Friday 7/31 is a listed expiry on all three indexes, so rather than scaling the weekly number down by the square root of time, I can just read the at-the-money straddle for exactly the stretch that’s left: ±1.43% on the S&P, ±1.90% on the Russell, ±2.45% on the Nasdaq, Wednesday through Friday. That’s a quote, not an approximation.

Measured against it — cushion being the distance from Tuesday’s close to the nearer edge of Sunday’s cone, divided by that remaining move:

• S&P — 1.08× to the top, 1.40× to the bottom. The roomy one, and only just.

• Russell — 0.79× to the top. Through the line.

• Nasdaq — 0.77× to the bottom. An ordinary remaining move, entirely inside what’s currently priced, takes out the low end of Sunday’s cone.

That 0.77× is not a forecast. It’s a statement that the Nasdaq’s cushion, on the side it’s been walking toward, is now smaller than the move the market is charging for the days that are left, with the Fed at 2:00 and four of the five biggest companies in the world reporting inside the next 36 hours.

A note on the method, because it changed the answer. Last week I sized the remaining move by scaling the weekly expected move by the square root of the sessions left. Doing it that way here would have printed 1.26×, 0.96× and 0.87×, and every one of those is too generous. Square-root scaling assumes vol is spread evenly across the term, and this week it plainly isn’t — the events are stacked in the back half, and the straddle that covers only those days is priced accordingly. I don’t know how often that gap is this wide. It’s wide enough this week that I’d rather quote the contract than infer it.

How far they went to get nowhere

Two sessions of distance against the cone drawn for five. Every bar clears its own cone width; the net moves are the thin markers.

Two sessions is a thin window and I’d hold the precision loosely — but the ratios aren’t marginal, and the direction is the same across three different indexes.

Efficiency is the part I keep looking at. Last Wednesday the Nasdaq was running 55%, which is one-way travel, a trend you can at least orient against. This week it’s 29%, and the S&P is at 10%: the tape covers a percent, hands most of it back, covers it again. That’s the regime that quietly bleeds a short-premium book through adjustments while the closing prints stay boring, and it’s what the market has chosen to do in the two sessions before the catalysts rather than after them.

What changed in vol

Same-method comparison — each gauge against Sunday’s read, nothing rebased.

The drift receipt. The back end fell hardest, which is new.

The IV ranks barely moved. Whatever the options market thinks about this afternoon, it decided it before Sunday and hasn’t changed its mind since — the Nasdaq is still the expensive lane at rank 70.7 and the 90th percentile of its own year, the S&P is still mid-range, the Russell still cheap.

The curve is the more interesting line. The 9-day-to-3-month gap has now flattened for a fourth straight read, 2.89 to 2.61. But the previous three weeks flattened from the front, near-dated vol rising toward the back. This one flattened from the back: VIX3M fell 0.65 while VIX9D fell 0.37. The market is marking down three-month risk on the day of an FOMC meeting that the rates market prices as a live 36% hike. Those two things can both be true, and they are, and it’s an odd pair to look at on the same screen.

A correction to Sunday’s lane table

Sunday’s issue printed a dash in the Russell’s variance-premium column and said its vol gauge was a backward-looking stand-in that shouldn’t be trusted. That was wrong, and I’d rather say so at the top of a section than in a footnote.

The Russell’s vol index, RVX, was never delisted. The data feed I pull from was returning “possibly delisted, no data found,” I took the feed at its word, and for weeks this newsletter has been telling you the Russell has no forward-looking vol gauge. CBOE has been publishing RVX daily the entire time. The puller now reads it straight off CBOE’s own feed, and every historical Russell row in the database has been rebuilt from the real index.

Here’s what that changes, using the same 7/24 close Sunday’s table was built on:

Sunday's whole argument was that IV rank and the variance premium disagree, and that the S&P's 79th-percentile cushion was the fattest on the board while the Nasdaq's 38th was the thinnest. The first half holds. The second half named the wrong lane.

Charged vs. delivered, per lane. The Russell column exists for the first time, and it's the tallest one on the board.

The Russell has the lowest IV rank on the board at 25.4 and the fattest cushion on the board at the 98th percentile of its own year. That is the disagreement Sunday was describing, in a louder version than Sunday could see, and it was sitting in the one column I’d been printing a dash in since June.

Two things I’m not going to dress up. Any Russell figure quoted in an issue published before this past Sunday came from the stand-in and does not match the corrected database — the 6/28 Setup said Russell IV rank 78 when the real RVX read 35, and that’s the largest gap of the bunch. And the Nasdaq’s cushion improving from the 38th percentile to the 52nd this week is mostly not good news: HV20 fell from 22.78 to 20.80 because a violent stretch rolled out of the back of the 20-day window, not because this week has been calm. This week it traveled 4.48% in two sessions.

The adjustment

Educational framing, not a trade — conditions, not signals.

• Distance is already spent; the events aren’t. Every index has covered more than its full-week cone width and none of the scheduled catalysts have fired. Whatever a short-premium position has already paid in adjustments this week, it paid before the part of the week it was sized for.

• The Nasdaq’s cushion is on the wrong side. 1.88% to the bottom of Sunday’s cone against the 2.45% the Wed–Fri straddle is charging. Inside the cone is not the same measurement as room inside the cone.

• Low efficiency is its own risk, and it doesn’t show up in a scorecard. 10% on the S&P, 29% on the Nasdaq. A week can finish dead inside every cone and still have been expensive to hold, and Friday’s Debrief will only be able to measure one of those two things from the closing prints.

• The Russell is now a lane you can actually price. Cheapest headline premium on the board, 98th-percentile cushion, and a real forward gauge behind both numbers for the first time in this newsletter’s life. Whether that combination is attractive depends on things this page doesn’t cover: small-cap liquidity, your own strike selection, what PCE does to the rate path Thursday morning.

• Anything expiring past 2:00 PM today is short the decision, and anything past tonight’s close is short Microsoft and Meta on top of it. Then Thursday morning brings PCE and GDP, and Thursday night brings Apple and Amazon. Map the expiry to the timestamps first.

I don’t know whether the drop in three-month vol is the market pricing a benign Fed or just the mechanical drag of two chopping sessions with no net move. Those are indistinguishable tonight. They stop being indistinguishable at 2:01.

What’s left on the week’s calendar

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Wed 7/29 — FOMC decision, 2:00 PM ET, press conference 2:30. No Summary of Economic Projections at this meeting, which historically leaves more of the reaction to the presser. Fed funds futures price roughly 64% hold / 36% a 25bp hike as of this week. Microsoft and Meta report after the close.

• Thu 7/30 — June PCE and advance Q2 GDP, 8:30 AM ET. Apple and Amazon report after the close, with Apple’s call at 5:00 PM ET. The densest day on the week’s map, and the one that decides most of what’s left of the cones.

• Fri 7/31 — Weekly expiry. Sunday’s cones settle against Friday’s close, and Friday’s Debrief writes the scorecard, including whether a week that spent its whole travel budget by Tuesday still managed to finish inside.

Your turn: All three indexes have traveled further than their entire weekly cone in two sessions, at 10–32% efficiency, with the Fed and four mega-caps still ahead. Does that read as the market getting the chop out of its system before the events, or as a tape that’s already unstable going into them? Reply — I read and answer every one, and the best reads make Friday’s Debrief.

— Theta Desk

Subscribe now

Subscribe now

TL;DR

Look at IV rank and this week reads simple. The Nasdaq sits at 69.4, the 89th percentile of its own year. The S&P sits at 29.1. One lane is paying you two and a half times the rank of the other, and the obvious conclusion is that the fat credit lives in tech.

Now measure the credit against what the index has actually been delivering. The Nasdaq charges 28.39 vol against 22.78 realized, a 1.25× variance premium, the 38th percentile of its own year. The S&P charges 18.58 against 10.41: 1.78×, the 79th percentile. The lane with the fattest headline premium has the thinnest actual margin on the board, and the boring lane has the fat one.

Then check the calendar, which is where this stops being an interesting table and starts being a sizing problem. The FOMC decides Wednesday, June PCE prints Thursday morning, and Microsoft, Meta, Apple and Amazon all report inside a 30-hour window — every one of those nodes lands inside the Nasdaq’s ±3.15% cone. The one thing to internalize: the index that cushions you least this week is the one carrying almost everything that can move it.

The regime right now

Vol went essentially nowhere week over week, which undersells what happened underneath it.

• VIX 18.58 — down 0.19 from last Sunday’s 18.77. Still the middle of the normal zone. If you only look at this number you would think last week was uneventful, and Friday’s Debrief is a 1,500-word argument that it wasn’t.

• Term structure: 17.62 → 18.58 → 20.51 (9-day → 30-day → 3-month). Contango, same as it’s been for a month. But the 9-day-to-3-month gap is 2.89 points, down from 3.69 last Sunday — and all of that compression came from the front. VIX9D lifted 0.77 while VIX3M moved 0.03.

  

  The curve, last Sunday vs. now. The 9-day gauge is the only part that moved — the gap to 3-month compressed 3.69 → 2.89, all of it from the front.

This is now the third Sunday in a row the front end has moved and the back end hasn’t. Two weeks ago the curve was steep as a wall and the message was “the risk is next week.” Last week it had flattened to 3.69 and the risk was Wednesday’s Tesla and Alphabet print. This week it’s flatter still, and the near-dated gauge is the only part of the curve that cares.

The receipt from last week, since we keep those: all three indexes closed inside their cones, and the Nasdaq traveled 8.39% of ground to net down 1.62% — 19% efficiency, nearly three times its cone width covered to end up roughly nowhere. A quiet close, a loud week. That’s the tape we’re walking into, and the front of the curve seems to have noticed even if the headline VIX hasn’t.

The lane table: which one actually pays?

Same-method comparison as always — IV rank against each index’s own past year, prior Sunday vs. now, plus what each index has really been delivering.

What each lane charges against what it has actually been delivering. The Nasdaq's headline premium is the biggest number on the board and the thinnest margin on it.

Two things worth reading off this table.

The Russell’s rank halving is not a repricing. It went 47.7 to 20.0, which looks like premium collapsing. It isn’t — its gauge is HV30, and what dropped out of that 30-day window was the early-July selloff. The small-cap options market may well be doing something else entirely; I can’t see it with this instrument, and I’d rather say that than dress the stand-in up as a reading.

IV rank and the variance premium are answering different questions, and this week they disagree loudly. IV rank asks “is this index’s implied vol high for this index?” The variance premium asks “is the implied vol high relative to the movement this index is actually producing?” The Nasdaq’s answer to the first question is yes, emphatically — 69 rank, 89th percentile. Its answer to the second is barely: 1.25× is the 38th percentile of its own year, meaning on nearly two-thirds of the days in the past twelve months, Nasdaq sellers were better compensated per unit of realized movement than they are today. The S&P’s 1.78× is the 79th. You are paid a bigger number in tech and cushioned less by it.

One caveat on that cushion, and it matters this week more than most: HV20 looks backward. The variance premium here is this week’s implied against last month’s delivered. It tells you what the market has been charging relative to what recently happened, not relative to what Wednesday and Thursday will do. A 38th-percentile cushion can be perfectly adequate if realized vol keeps falling from here, and can be nothing at all if it doesn’t. I don’t know which. That’s the honest version.

Expected moves this week

The expected move is the market’s own one-standard-deviation estimate — roughly a 2-in-3 chance of finishing inside the range. These come from the actual straddle on each expiry, so unlike VIX they price this specific week, catalysts and all.

This week's ±1σ cones through Friday 7/31, with the week's event nodes marked. Every catalyst sits inside every cone — and the widest cone has the thinnest cushion behind it.

• S&P 500 — spot 7,411.98, weekly ±131.6 (1.78%), range 7,280 – 7,544. Daily ±52.0 (0.70%).

• Russell 2000 — spot 2,930.0, weekly ±68.4 (2.33%), range 2,862 – 2,998. Daily ±28.7 (0.98%).

• Nasdaq-100 — spot 28,128.34, weekly ±886.2 (3.15%), range 27,242 – 29,015. Daily ±364.3 (1.30%).

The part I found genuinely interesting: same five-session horizon as last week, and every weekly cone got wider. The S&P went from ±1.57% to ±1.78%, the Nasdaq from ±2.91% to ±3.15%. Meanwhile the 30-day gauges that supposedly measure the same thing went down. VXN fell 0.64 points and the Nasdaq’s cone widened by a quarter of a percent.

Those two readings measure different instruments. VXN is a rolling 30-day blend. The weekly straddle is a price on the specific five sessions that contain the Fed, PCE, and four mega-cap prints. The straddle charges up for those five sessions; the 30-day blend dilutes them into three weeks of ordinary tape. That’s roughly what you’d want the options market to do, and it’s a reminder that a 30-day vol index is a poor proxy for the week you’re actually selling.

What it means for premium sellers

Educational framing, not a trade — conditions, not signals.

• The Nasdaq is where the credit is and where the cushion isn’t. Rank 69, percentile 89, and a 38th-percentile variance premium. That combination is specific: it means implied vol is elevated in absolute terms and the index has been moving nearly enough to justify it. You aren’t finding a mispricing there this week so much as accepting a fair-to-thin price for a genuinely eventful five days.

• The Fed is Wednesday afternoon and the mega-caps are Wednesday and Thursday night. Any expiry that lives past Wednesday’s 2 PM decision is short that event; anything past Thursday’s close is short Apple and Amazon on top of Microsoft and Meta the night before. Map your expiry to those two timestamps before you map it to the credit.

• A wider cone is not a safer sale, it’s a bigger invoice. The S&P’s weekly cone widened 13% week over week. You collect more because the market expects more, and it expects more for reasons printed on a calendar.

• The S&P is the cushioned lane this week, which is not the same as the safe one. A 79th-percentile variance premium is the best margin on the board, and the S&P still has the Fed and PCE inside its ±1.78% range. Better cushioned, same macro.

The trap this week is the mirror of the one two weeks ago. Then, cheap premium tempted you to add size to make an income target work. This week, the highest rank on the board tempts you to read “69th percentile” as “well paid” without ever checking the second number — and the second number says the Nasdaq is paying you less per unit of movement than it did on most days this year.

The risk lens

What a careful seller actually watches this week:

• Wednesday 2:00 PM ET is a scheduled discontinuity. The FOMC statement lands, the press conference follows at 2:30, and there’s no updated dot plot at this meeting, which historically leaves more of the reaction to the presser’s tone. Two hours of cash session left afterward, and short-dated positions carry all of it.

• Thursday morning is a data node and Thursday night is an earnings node. June PCE and the advance Q2 GDP print both hit at 8:30 AM; Apple and Amazon report after the close. If you’re short premium in NDX, you’re short both ends of that day.

• Four names, one cone. Microsoft, Meta, Apple and Amazon are the heaviest cluster of single-name event risk the Nasdaq sees in any quarter, and all four land inside a ±3.15% weekly range on a 38th-percentile cushion. Whether that cone is wide enough is the whole question, and nothing on this page answers it.

• Don’t size the Russell off its own gauge. Its IV reading is a backward-looking stand-in that just shed a volatile month. A number that mechanically falls when the past gets quieter should not be the input that makes your small-cap position bigger.

• Last week’s travel is the argument for sizing to the path. The Nasdaq covered 8.39% of ground to go 1.62%. Put a week with four mega-cap gaps and a Fed decision through that same behavior and the endpoint may still land inside the cone — the ride to it is what you have to fund.

Watching this week

Standard release windows below — confirm dates and times against your own economic calendar before relying on them.

• Mon 7/27 — Quiet on the macro side. The week’s weight is Wednesday onward.

• Tue 7/28 — FOMC meeting day one, no statement. July Consumer Confidence in the morning.

• Wed 7/29 — FOMC decision, 2:00 PM ET, press conference 2:30. No Summary of Economic Projections at this meeting. Microsoft and Meta report after the close.

• Thu 7/30 — June core PCE and advance Q2 GDP, 8:30 AM ET. Bank of England decision earlier in the morning. Apple and Amazon report after the close. The densest single day on the week’s map.

• Fri 7/31 — Weekly options expiry; this week’s cones settle against Friday’s close, with Thursday night’s two prints digested in the cash session first.

The premium looks richest exactly where the cushion is thinnest, and that’s also where the entire week’s calendar is parked. Being paid a big number and being paid enough are two different measurements, and this week they point in opposite directions.

Your turn: The Nasdaq’s IV rank is 69 and its variance premium is the 38th percentile of its own year, with the Fed and four mega-caps inside its ±3.15% cone. Do you take the bigger credit and accept the thinner cushion, or sell the S&P’s 79th-percentile margin and give up the headline number? Reply — I read and answer every one, and the best reads make it into Wednesday’s Recalibration.

— Theta Desk

Subscribe now