Subscribe now

A confluence of different implied things would all point to the same outcome.

The 6.8 Extension

SPX is at a massive technical level at the 6.8 extension of the 2007 - 2009 swing.

A full free 7 part series on this and the different bull and bear outcomes has been done, covering the historic evidence for the claim of this being an important level, showing different types of bear moves and also the bull moves (which are hyper aggressive, this is the bear’s last real level for this to happen).

In the 6.8 rejection setups the minimum expected drop is 40%. Extremes of 6.8 rejection happened in 1929. This broke the supports and end up down over 80%.

So in the event that the 6.8 is a resistance this would give a best case 40% drop.

Read the 6.8 research.

Semis

The semis charts are extremely alarming - and I do try my best not to be an alarmist. They’re bad.

It’s not that they are down, it’s the way they are down and the current patterns they are in are common precursors to capitulation.

These are patterns that in late development. Very close to their trigger points. Currently most of the semis charts show patterns where if something does not change dramatically they look set to make breaks that are almost never expressed in anything other than crash moves.

These technical breaks are most often news driven.

Carry Trade

The Yen is at extremely long term technical supports. This is the case on the index and it is also the case on the majors and the pairs used for carry trades.

While SPX is at the super long term resistance, and while the semis are showing pre trend failure patterns the Yen is sitting at extremely big supports with central banks seeming actively interested in the supporting of the Yen. The Yen trade unwinding would be a big drain of capital from the other markets.

Euphoria Inducement

A classic thing found before a break comes is there being sudden bursts of extreme moves in things that were under performing recently. Giving a “Now even that is rallying, it must all go” vibe. We have this in record one day gain for a SPX in MNRA and recent crypto pops.

The recent pop in BTC being modellable inside of a bull trap.

AI Overbuild

The profit potential of AI could be egregiously mis estimated, or at least in terms of how it will relate to the market of the coming years.

We may be in the late stages of a tech bubble, having many markers similar to the Dotcom bust and being passed most of the KPIs of that bubble in terms of valuation metrics.

Share

Subscribe now

Additional context on why the 6.8 is important and these rejection paths are historically valid is in the free series linked below;

Full series, or see part 1 below.

I wont spend too long on the 6.8 here because the thesis is simple and it’d be reputative to go through it all again. The important bit is the 6.8 is historically a major trend decision point. When the trend decision is downwards, there are two main types of patterns that happen.

One is a “healthy correction”. It’s big and nasty. Price travels from the 6.8 to close to 4.23. 4.23 does not usually hit and in most instances this means the asset in question drops 40%. This is a great retest of the trend and when we make a new high above the 6.8 this has most often been an important breakout - a notable failure of this ideal was the 2007 breakout of the 6.8 which would later crash to 4.23. It would then have the uptrend properties on the rebreaking of the 6.8, but you cant really gloss over a false signal into a crash high.

Instances where the 4.23 is broken usually seen bull traps off the 3xx fibs, spiked above the 4.23 but then broke down later and these were overwhelmingly found in instances of full trend reversal (1929 - 1930 break being the first one).

Crash Studies

My method does not allow me to map out the concrete plans I’d expect in a crash in the same way I can in a rally because with the rally forecasts the anchor swing for the fibs is known. It’s the 2007-2009 move. That’s not changing. But for the downtrend pattern I need to use the topping swing of the uptrend. If I do that today, it may no longer be true tomorrow.

Instead what I want to focus on is explaining the methodology for planning important levels in a crash/correction. Because the broad 4.23/6.8 rules are useful but for practical trading is riddled with caveats and problems. So what I want to show you instead is how to use the topping swing to map out decision levels and how a crash or recovery tends to look around those levels.

This will give you a functional way to assess the trend/correction/crash “health” and important levels in any future downturn should there be one.

I’ll end with my guess at where the high may be and a crash forecast for fun, but the method is the important thing. I want you to be able to do this yourself if it happens.

Why its a Bubble

I am fully convinced we are in a bubble and I just dont know what part. The TA is neutral but with extreme forecasts to either side and simply saying “AI” is enough to support a case for us being mid bubble or end bubble.

Here’s the case for end bubble. Briefly. This is just rhetoric. It doesnt matter and you can skip to the crash study for the useful TA strats.

LLM agents are expensive to run. They use a lot more compute than expected. This is electricity costs and it means a wall has been hit on this. Compute isnt something that can be easily solved. Its not a software problem. Real things in the real world have to change to accommodate it.

Outside of these agents, inside a strong harness and specifically tooled the economic value of LLMs is close to nothing. I mean ... it’s really not much at all. The people who are benefiting from these are small teams/solo operators and they’re usually making money for themselves and will never employ anyone. It’s good for people using it well but its not widely valuable.

LLMs are a massive liability to businesses. A headline court case was when a guy asked a LLM chatbot on an airports website about a discount they could claim. The bot said they could and just to do it when they got back and had time. Policy was it had to be done the spot. Claim was denied. Court case. Airport lost. “If its your bot, you said it”.

And thats a pretty sketchy thing. A misconfigured bot is at the very least bad for brand image but when its also getting you sued that makes it far less attractive to use these.

OpenAI and Antrophic are the only real big users of the compute. Outside of them there isnt all that much business and these companies are not making money. Antropic has a smarter business model because they probably do make money on the API which they have big deals with - but both of these companies (probably) lose money on their $20 - $200 a month subs.

The shiny King that supports “Ah but AI has earnings” is NVDA. And NVDA is just borrowing money to give to people to buy their things. Then build data centers and put their things in. And then NVDA (directly or my proxy) will rent those things off of them. This is creating a lot of transactions. These are being counted by different companies as earnings - but if 5 of us sit in a room and pass around $10 to each other, did we get richer?

A bear thesis would have it that we are inside of a massive infrastructure overbuild. It’s fuelled by claims of capbilities that can not currently be done good enough, reliabily enough and cheap enough (has to be all these). You can in many cases get something that is one or two of these, but all three is hard for most use cases and would be heavily dependent on the harness and tools.

And sometimes they say NVDA is Enron.

The 140 Year DJI Study

Since its inception the DJI has dropped over 10% 42 times. Of those 42 times, 38 were suitable for the methology I am going to show you. Only 2 of these failed the expectations with all the others acting reasonably as expected per the model / norms I’ll propose.

I wont go through all of these because it is 38 events. It would become a redundantly long post but I will go through a lot of them and I’ll go through 100% of the major ones. I want to show you a lot of examples so you can learn the methodolgy and you can see that I actually did this. I didnt ask a LLM to give me my opinion. I went through every event. Build sims to test and interrogate my work and found it to be robust. Which is why I am sharing it now.

For the pre Roaring 20s we will take the 21 Depression. Two reasons. It was the last crash before the mega rally of the Roaring 20s started. Two, you may not know that this was classed as a Depression because it was so brutal. Lost to history by the far worse one that came off the rally this one setup.

One a topping swing is made we draw the fibs low to high of that.

You can read about how to draw topping swings here.

The typical rules are price will make a low 1.27 - 1.61 in a strong uptrend. Breaks under 1.61 usually lead to capitulations to 2.20 - 2.61. These levels can be lows but stronger selling under them. 1.27 or 1.61 is often retested before bigger crashes. With only one exception (depression) crashes end around the 4.23 fib.

This one made the first break and then held the 2.61. In the left panel you can see info on the fibs held by other crashes in that era.

The 1.61 was found repeatedly to be a strong deciding level between a small correction and more aggressive pullback or even crash. In strong trends the 1.61 was the optimal level to look for lows. Here’s the 1906 SF Earthquake.

In the early era the worst crashes were 3.20- 3.60. With World War One being an example of one.

The Depression is the only instance of a meaningful 4.23 break and that made a low at the 6.8.

The 1937 crash was the first time we seen a 4.23 extension. Those candles you see coming off the 4.23 low could probably be reasonably named the real end of the Depression. The low had been made by then, it had drifted up but this was the first time it made a major sell off, failed to make a new low, went onto make a new high and was the failure of the downtrend.

Heading into the next era, obviously 87 is the most famous one. It’s a 4.23.

The 1962 flash crash is also interesting. In these “flash” events this is when you’d expect levels to be meaningless but you can see even in this strong selling all the stalls are at fibs and then the low is at one.

Into the next era and the famous 2008 crash. This one is very picturesque and I already knew this one before doing the full study. Its what my belief crashes could be modelled was mainly based on.

Classic interactions all the way down. 4.23 low.

Subscribe now

I guess all of them in this era are famous so lets do them all. 2000 is a 4.23. Damn ugly crash. Yuck. Would not have been fun to be bullish with those deep bull traps.

The 2020 one. Again, flash crash. All news. Should not make any sense. 4.23 low with path interactions.

2022 was a 2.61. I posted about the case for the end of the bear market and new highs on this one here in real time.

Could new crashes sneaked in since I did the study. Quite rude. I’ll do them manually to be comprehensive.

2025 is 4.23.

Very recent one is 2.20.

In terms of the two that failed, they both were shallow corrections and both failed in the same way by making a low in the middle of the 1.27 - 1.61, which the method would class as entirely random for a low. It should be 1.27 or hit 1.61. They really were quite minor moves with the norms holding up in most shallow corrections and all real breaks.

They were so small I cant even remember what ones they were, but they looked like this.

Subscribe now

The proposal here is when you are able to confirm a topping swing, the downside extensions that gives are useful for crash/correction decision levels. 1.61, 2.61 and 4.23 are the most important binary inflection levels but all fibs have represented lows. This can be used for dip buying levels if bullish and trailing stops if breaks.

Inside of bigger crash moves, the bull traps often retested broken fibs and then dumped. Useful for selling dips, trailing stops on longs from lows and being cautious about adding into strong local rallies at critical macro resistances.

They do not, in any way, help forecast the high. The usefulness of the method is entirely based on the fib being drawn from the high swing which can not be known before the fact. The strategy comes into play when the 1.27 of the topping swing has hit. This will, by definition, be a structural break of the trend also.

Its also important to note that since the downside levels relate to the size of the topping swing a crash being 4.23 does not have to mean its the biggest. 2020 and 2008 were 4.23 crashes but different percentages. So this does not (until top swng is known) have any way of forecasting the depth of the crash.

==============================================

With all that said ....

Here’s my guess at what it would probably look like.

Nicer one: Top 8000 - 8250 with a low somewhere 4600 - 5000. More on the lower end of that band.

Nastier one would require a bigger topping swing and I’d guess again around the same 8000 - 8250 high but this would give a depression style target on the 4.23 at close to 1000.

These are covering the absolute extremes of the shallow pullbacks or major reversals in the biggest assets on the biggest timeframes. The same ideals can be used for most assets and timeframes when it comes to planning correction levels and levels where a correction has become a reversal.

I’ve covered this all with reversals to upside moves but this also works in downtrends. I learned this trading Forex intraday downtrends. Turns out it tests well against the main events too.

Read Full Supporting Series for the 6.8 Decision Level.

My technical works is far more reseached and plans contingencies better, you can read it free here:

Read First:

The Case for Massive SPX Decision Soon (Part1)

The Case for Massive SPX Decision Soon (Part2)

The Case for Massive SPX Decision Soon (Part3)

(Ongoing series).

Firstly I want to say although I do not think LLMs are going to prove to be “Good enough”, I personally have found them amazing.

When it comes to the things you can do with them, they are incredible. I started to play with LLMs on ChatGPT 3.5 launch and continue to use them daily today. I am not in a position where I have 10 times more than I could have thought of doing before LLMs. But I think they are missing the needed mark.

From the earliest versions of LLMs I could see these would be useful to me and they actually made it very easy for me to believe we may head into a full blown bubble. The productivity gains implied were huge but this was built on three core assumptions. LLMs will keep getting better, tooling will improve and costs will drop.

One and two were correct and what you can do with the most recent models inside of the best tool harnesses is amazing now. But they are expensive.

The really strong utility in LLMs comes from agents. All the things that support these massive investment in them relate mainly to agents rather than simple LLMs.

Simple LLMs are good for things like summaries, covering notes to full prose, proofreading, writing emails - its not trillion dollar stuff.

When it comes to things of real economic value long running agentic sessions are needed. Without getting too into the weeds on how agents work the only important thing to take away is when you give an agent one prompt it can keep feeding its own output back into the model, and keep working on a task.

Inside of well designed systems this makes a lot possible, but its expensive.

Subscriptions Subsidised

When you buy a $20 to $200 sub from the main providers like OpenAI or Atrophic you get a lot more than what the API price to buy the same tokens are.

I’ve read other’s work into this and see things stated like one full max limit user needs 20 dormant users to cover it inside of heavy coding use. Seen people cite things like if you use more than 25% of your monthly limits you are unprofitable to the upstream company and other things like this.

I didn’t fact check this but I can tell you antidotally from my own testing you can burn through $100 in Claude tokens in about the same amount of work as you can do in a day to a day in and half on the $20 plan. You can get away with this heavy usage for about 5 days out of the week before hitting weekly limit. So let’s say that’s 3* as much.

For what you’d use at max usage on the $20 plan, you’d spend $300 on the API.

The Compute Crunch

Because of the nature of how agents work they are more compute intensive. If you discard any rate limits its entirely viable you can write a prompt of 100 or so word into a coding tool and walk away with that running for hours on end and using up 300K - 500K tokens. For context, books are about a million tokens worth of words.

This has turned out to be far more costly than the upsteam providers expected and this is where I think LLMs are running into the issoe of not being profitable to provide. There was an insanely big signal of this recently when many companies started to do things to reduce their compute problems.

“Github” is one of the biggest sites online when it comes to coding stuff. It’s owned by Microsoft. You may be aware, Microsoft has a few assets here and there when it comes to compute … And Github had to entirely suspend signing up for new paid accounts because they had to work out how to make the compute viable.

You could still sign up free, you could not give them money to use the smarter model in agent mode.

What does it mean when the company with arguably the best assets in the world to deal with this have to pause their PAID product?

There are other examples of this, Amazon’s agents now have token caps on them. “AWS” is the provider much of the internet runs on and Amazon are known for their tolerance to losses to gain market share - but it seems even they do not like the costs of this.

Costs Vrs Income

The cost for all this is huge. Even if we ignore things like energy and such as in “Costs to the world” it’s still incredible expensive to be building these datacentres to manage all this. But is the actual income available from this isn’t clear. Even at enterprise level where the costs are usually highest its hard to see how all this money is getting made back.

Let me frame it like this, Sam Altman said something to the effect of, “We made a soft promise to our investors that we will try to create AGI and then once we have we will ask it how to get them a return on their investment” - meaning, “Because we have no fucking ideas!”

By Altman’s statement the bet on LLMs is now that LLMs will become AGI. Whether they actually can or not in an area of deep debate. If AGI would be a good thing is even hotter debate (seems like a bad idea to me, I hope its not possible) but the crux of this is its a super speculative statement.

Now we have recently started to get talk of these big upstream companies starting to talk about “Pacing” LLM development. This is being presented inside of “For safety”. A cynic could read it as they are hitting a wall and make out they have to slow down rather than ran out of steam is a better way to deal with this.

Either way, it doesnt go well with the “We’ll ask the super AI” statement if work on the super AI is to be deferred and regulated.

This comes off the back of different reports about LLM safety. The “Fable” model was briefly taken offline for non-US users, apparently due to fears of cyber attacks or bio-weapons. Then OpenAI said one of their unreleased models broke out of its sandbox and then hacked hugging face (biggest community for LLM stuff).

Anthropic then said, “Oh … we checked our logs and turns out our model broke out and hacked things on 3 occasions, too”.

The premise of LLMs is they need to be good enough, reliable enough, cheap enough and safe enough - and these things are not coming together.

On the models we can use if they are good enough, they are not cheap. And since this is tied directly to the compute crunch this is unlikely to change soon.

Even if they are good and they are cheap enough, they are not safe. Or, they are being given to us by lunatics who can not be trusted with them.

Think about it, on the OpenAI thing, either the model was so incredibly good that OpenAI had all the protections anyone could imagine, but the model broke them. Then after it did, it went on to commit a series of real world crimes and all of this was done without observation (these were found in the logs after it happened).

Or, OpenAI have something that has the potential to hack major sites and potentially do destructive actions and they didn’t bother to do the things to make it safe.

And if it’s not one of those, it’s a bunch of BS … it’s marketing. “Look how super strong our model is” and if that’s happening, why? Did they hit a wall?

Tooling and Infrastructure is Essential

To be able to trust these LLMs on any super high stakes venture (like say, running your business) the LLMs themselves should not be trusted. Even at the top end of what is possible the smart use for LLMs is to build tools and workflows that allow you to use the good bits of the LLM but not allow it to do anything risky.

This is entirely viable and when we are talking about LLMs inside of this use case there are a lot of things they are absolutely “Good enough” for. They are plenty good enough for the things I want to do but this is a small, contained, use case and it still has taken me a year of planning, testing and improving to get it ready.

Even now, I can only be really sure there wont be mistakes in the most critical of things that I’ve built the best tools for. Everything else is ongoing checking and tool improvement - and this makes the idea you can plug a LLM into your business and reduce your staff without much thought crazy. It’s not viable.

The backend work to make it viable is a lot to do. Now … this is actually something LLMs are very good at doing. They can be used to build this tooling for them to use, but when doing this right the best way to remove the most risk also tends to be work that makes it so the model you use can be faster/dumber.

That may sound good because it solves the upstream compute problem but the issue is if I can build a system where I can use the newest Claude / OpenAI models or I can use a couple year old Deepseek model for 95% cheaper … how is the money being made back?

It seems to me people dont have the understanding to make LLMs safe/vaiable for business ops nad if they do, they have no reason to use the major models most of the time.

Indeed, a wise way to do it would be to subscribe to Claude Code at $200 a month, burn up about $3,000 a month in compute costs to build this fancy backend system and then switch to a deepseek model giving you 300 - 500 requests per $1 spent.

The only way this seems to work out is if things get really good, really fast. As someone who uses all the models on a regular basis and for tough tasks I can tell you they do get better. Perhaps not in ways obvious from causla chats or use cases but they are always getting better- but they are not entirely reliable.

Now, here’s the thing. The whole AI narrative is built on “AI is making real money, companies have real profits” … but they dont.

NVDA has profits selling things to a small number of hyperscalers. Those hyperscalers are bleeding out cash. NVDA is now making proposals to pend other companies $300 billion so they can make data centers, buy NVDA chips and then NVDA are going to rent them back from them.

So is it the case that NVDA can “Make money” as long as they lend that money to people to buy their products and then buy their own product back off them?

I’m not a fundies expert but that seems a bit like “Not making money”.

Few practical caveats;

• I might be wrong about all or most of the above. It’s opinions based on observations and not deep research.

• Some huge LLM development could change all of this.

• “AI” is not only LLMs. LLMs are doing all the headline work right now but other “AI” might come into play and change this whole thesis.

• All of this could be right and the market goes insanely parabolic. Why? Who needs a why?

My technical works is far more reseached and plans contingencies better, you can read it free here:

Read First:

The Case for Massive SPX Decision Soon (Part1)

The Case for Massive SPX Decision Soon (Part2)

The Case for Massive SPX Decision Soon (Part3)

Subscribe now

Join our main newsletter for more regular updates and analysis.

Please read this free post first. We won’t be repeating thing for context.

In the previous post, we used the 4.23 extension of 2008 to show how we have similar situation to here.

In recent forecasts we’ve said the local action looks like it might setup a wave 5 spike low. Which would look like this.

Here’s the interesting bit. Let’s look closer at the DJI on the daily timeframe.

If we were in the final section before a squeeze top - we’d be in here.

Range and head fake. Looks similar, right?

And local analysis tells us we might see a sharp spike down and then that transition into a rally. We are forecasting a nominal break of the low in the drop.

If that hits, it’ll look like this candle.

In case you do not know how to read these old school bar charts, this is telling us the market gapped down. Traded up (Probably instantly) and then closed at a double bottom. Would have looked bearish AF(!!!) but this setup the final squeeze high.

Would imply these two swings coming.

This analysis is currently just a bookmark. Putting it on your radar.

If and when we have swings that further imply this may happen and we’re in zones we can define smart bets with good stops - we might look at some trades for it.

First thing we have to see is the big false breakout and bear mania.

Bears should feel confident before a squeeze starts. That should be achieved by a bull trap turning into capitulation.

The next rally will then be a strong and true one - but by this time people will have been conditioned to sell the rip.

Another big spike high. Huge failure of bear forecasts. Bull mania. Then … maybe .. it’s time to try for a legit reversal trade.

Subscribe now

For more regular updates and trade plans join:

Please note, “Close” is a relative term here. We could still be 10% or so from the high and if we are, this can be the most aggressive part of the move.

But we’re in an area where there’s a lot of risk.

Let’s start with the biggest perspective.

Multi Decade Resistance

Heading into the 1920’s rally, DJI made a 50% crash. The rally would ultimately go on to be a 4.23 extension of this crash.

It chopped around there for a while. Spiked out, and then the crash had begun.

Here’s a picture from close to the 2022 high. When we first hit the 4.23.

This is how we look now.

Here’s another 4.23 top. This one from BB in 2021.

The BB spike resolved like this.

Which didn’t surprise me. I’ve been setting targets and trading reversals from 4.23s for years.

But have never seen one this big in the wild.

The previous fibs mark out the major highs and lows in the market since them.

The 4.23 rejection would be the biggest trade these fibs would have predicted so far, and they’d have predicted the high and low points of 2020 and 2022.

A full 4.23 rejection trade would be disastrous. The 4.23 mean reversion path is a drop to the 2.61 followed by a drop to the 1.27.

Here’s the OG.

And the percentage drop that would forecast would be the same type of percentage drop seen in DJI 1929, Nikkei 1990 and Nasdaq 20000.

Here are all of those together.

Were one to express the implied forecast of this model with a mountaineering analogy, we’d on the summit advance now. Which is often a little mountain on top of a mountain.

The zone they call the “Death zone” on the major peaks.

If we’re not at the actual top, but we’re getting to the top - then the final accent is going to be the steepest part of it.

I’ve been speaking for months about the style in which a final short squeeze forms.

But, if this thesis is correct, in the grand scheme of things we’ve seen best of this rally now.

Anyone calling for a bubble top here will look like a genius in ten years. As traders, the devil’s in the detail. We’re not here to look clever, we’re here to make money.

If we head into a squeeze it’s possible we see day after day of relentless rallying. In the same style we’ve seen RUT and DJI moving recently.

There’s a difference between bear market rhetoric and bear market strategy. The moment before the perfect moment is the worst moment.

But I do feel some strategic bets on a crash coming up. I don’t think SPX has topped quite yet, but I do now believe we may be late inside of the topping process.

More detailed plans on how we’ll deal with a squeeze and can we can begin to plan, and position for, a jackpot bear attempt in the near(ish) future.

I think we must be getting close to something big. Perhaps something that would bring about the fat tail risk previously discussed.

Approaching a Fat Tail Event - by HoleyProfit (substack.com)

All the patterns are bearish. So many big bear harmonic patterns.

Something happens when big harmonics form! The implied signals here are capitulation (Not throwing out buzz words here, I use the word selectively).

Pre-Capitulation Conditions are in Place - by HoleyProfit (substack.com)

Failure of the harmonics would likely start an intense squeeze. You know those intraday rallies you’ve been enjoying, bears? Yeah … imagine a real one of those.

My bias comes down so incredibly hard on the bear side. For so many reasons. But also if this confluence of setups fail I think I’ll have to be a new high guy.

Something big seems incredibly likely soon. The only question is how long do we range and I think perhaps we’ve whipped out everyone and it might be coming sooner rather than later.

Very strong spot to speculate on moves larger than the recent average.

VERY BAD SPOT TO GET CAUGHT OUT!

One sided momentum very likely once this decision is made.

It’s a sequential trade plan. Which means as one swing hits it supports betting on the next swing in the pattern. If swings do not hit - it was wrong is no longer useful.

Full Crash Forecast:

Subscribers, go to this version. It has everything in the current post in it but it also explains how the analysis is done and offers trading notes for each section of the move if it develops.

Market Forecasts and Trade Plans is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

A Full Crash Roadmap With Trading Notes. - by HoleyProfit (substack.com)

Swing 1: The first Break (Bearish)

A move down to 3950. This being an intermediate low and there being a whipsaw move starting in this area. Off a level just under 3950, we see some sort of tricky candle with wicks on both ends. .

Swing 2: Whipsaws (Bullish into bearish)

Swing 2 is a chop swing. The wicky candle spoken of above. Low is around 3940 and high is around 4140. Price is liable to whipsaw very hard in this zone.

Swing 3: The Fear Stage (Bearish)

At this point, the market goes into a crash relative to local timeframes. This crash takes us to retest the lows of 2022. Swing high 4140 areas and swing low 3550 area.

Swing 4: Two Legged Bull Trap (Bullish)

From here, another rally is setting up but this will be whippier. Approximate first bounce to 3770. Sharp drop for double bottom. Big bull trap rally to 3950.

Swing 5: The Next Bear Break (Bearish)

Downtrend resumes but has a few tricks to play. We drop from 3950 to a new low but quickly rally back to 3750 area. Probably a really strong local chart rally after the false bear break.

Swing 6: Prime Short Setup (Bearish)

Local chart crash event from 3750 areas to 3410 areas. Probably a really easy leg to be short in but few will want to be after the false bear break and vicious counter rally.

Swing 7: Fast Retest of Break (Bullish)

Another vicious bull trap. Main swing heads back to 3580. Possible a false start on the short here and a head fake above to 3650 areas. Really good short op into the head fake.

Swing 8: Near Term Capitulation (Bearish)

A crash of a bit over 10%. Maybe close to 15%. High around 3650 and a low somewhere close to 3150.

Swing 9: Big Stop Hunt / Bull Trap (Bullish)

Another bull trap. Probably made up of one really strong move and then a head-fake high. Low around 3150 and high around 3500.

Swing 10: Pre-Crash Break (Bearish)

A bear swing that makes a double false breakout at the low. Turns into a sharp bull trap. High around 3500. First low around 2970 and second low around 2900. Bull trap from 2900 likely very aggressive.

Swing 11: Last Big Bull Trap (Bullish)

A bull trap and one that is a total layup for a great short. The bull trap should have a lot of momentum and head to around 3420. Some sort of mini spike out at the top and the best short of the whole move is setting up.

Swing 12: Headline Making Crash (Bearish)

The Event. Without making any guesses as to what the event may be, at this point to trigger the downswing there will probably be a fundie trigger. The fundie supported crash starts here. 3420 to 2420.

Swing 13: Start of End of Drop Chop (Bullish)

The crash is close to over. Swing 13 will be a false hope rally and the next bear break will terrify people but the crash is close to over now. One more big bull trap. Low around 2420 and high around 2870.

Swing 14: Penultimate Bear Move (Bearish)

The first of two terminal swings of the bear market. Swings 14, 15 and 16 make up an AB=CD type pattern with it being one crash to a new low, a little rally and then the bleed out into the real low. Swing 14 high 2870 and low 2330.

Swing 15: Final Failed Rally (Bullish)

Last bull trap. After this, buying the dip is expected to beat selling the rip but by this time dip buying is a thing for museums. Low 2330 and high 2800.

Swing 16: The Bear Market Low (Bearish)

The kick out. A news supported move into yet another low. Despair takes hold at this point and it's the time when people who swear they'd never sell their assets now will be forced into doing so at the low. High 2800 and low around 2000 with maybe a head fake under there.

Swing 17+: The Turn to Uptrend (Bullish)

And now the crash is over. There will still be at least one more crash-like move but a new low will fail and we'll be in an uptrend. Although the crash would now be over, it would be a very (Very) long time before the recovery was made. The insta-recovery people have become accustomed to would not be applicable in this type of crash.

Market Forecasts and Trade Plans is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

I am currently long with stops trailed. This looks more like a bull break, but this is a good spot for bears in terms of RR.

If you want to speculate on a big drop or make cheap hedges, this is the ideal spot to do it. It’s possible if we’re in a turn this is the end of the bull trap and also we’re very close to where you could get out hedges into a break if you wanted to. If this works for bears, a big drop would likely and OTM puts would work.

We’re now retesting the 76.

If the action above the 76 was only a false breakout, this is when we’re likely to see evidence of that by 76 not being able to break a second time. This would be the optimal time to short in a bear setup.

Bulls traps in these two legged form set up big breaks. If this turns into a bull trap it’s very likely the move down is strong.

We usually can draw a fib low to high of this and look for the important levels to the fibs generated from that.

Moves like this would be an extreme warning for bulls! We should be seeing the bulls be strong now if they’re going to break. If there a big dump that should stop around the 1.61. At this point we can say definitively if there’s a break of this 1.61 our bias is much better to be shorting rips.

If it is going to happen for bears, I think it would happen here. Sellers taking over at this level would have me far more interested in making more detailed bear plans.

Bulls breaking this level I think is a string hint we’re going to over 4900.

• I am currently long and have stops trailed up to avoid getting rugged.

• This is probably the most supportable bear level (4500) of the year (But in my view it maybe failed already).

• Holding this level (4500) would be good for bears/bad for bulls.

• Likely an aggressive bull on a break here.

• Likely a rapid retreat from bull plans if hold.

• Above 4555 and above 4600 strong signs of bear failure.

But what’s the historic reality? Well, I checked. The odds of this being right are not as good as a coin flip.

Historically out of five occasions rates were raised at very best you could do for bears is give them a 20% chance. The maximum drop in SPX on interest rate rises is historically 17%. The maximum gain is historically double that at around 35%. When SPX dropped 18% from March 2022 to the low - this completed the maximum expected move on rising interest rates as per the historic norms.

I didn’t bother to check this before because it’s not relevant. My TA forecasts of a bear were posted in 2021 and the interest rates stuff came in 2022. When I said long USD in 2021 the only thing people spoke about was inflation. So checking this didn’t matter to me. I already had my plans.

When I was revering long I did the most basic due diligence on interest rates. A lot of people were telling me I was doing something very foolish and just to cover my butt I checked. But all I checked was time from interest rate hikes to a market high and I checked for the dates I know there are significant market highs.

I bought into the narrative interest rates drop markets and I was just curious as to what sort of time delay was possible. Of course, as my findings at the time suggest and as markets have proven since, stocks can rally on interest rate increases. Often making new highs, my research found.

But it was only when I went through the systematic backtest of every single possible bull trap over the last 80 yrs something struck me. That things was, the 70s were actually quite uptrending. I’d never paid much attention to the 70s because it’s flat on the monthly chart - but daily chart kinda uptrends.

I was happy about this because a few times I’ve wondered if after hyper trending markets we’d have a long slow range like the 70s in my life time and if that’d be bad for trading but it turns out the 70s were epic. Lots of big ups, lots of big downs. I’d have enjoyed them.

But then I thought something else. Something I really should have thought about before. I thought, “Didn’t something happen in the 70s?” Something to do with … interest rates. Actually, wasn’t the 70s the time when interest rates went from 5% to 20%?

And it was. And by this point I’d just looked through the entire 70s and there is notably no major crash. There are various big corrections but the 70s there’s no big crash. It was a really strong buy the dip market.

This is what happened.

And this really caught my attention because in the bull trap analysis post I’d happened to pick out this exact sequence of price swings to show how we have similar price swings in the current SPX chart.

So here’s the situation. The last time the FED did these things with interest rates and charts looked like this, there was no notable bear market and price was flat at best. Overall it was more of an uptrend. This is while we had the exact same economic outlook, at least in terms of interest rates.

That’s the actual facts of the matter. For all the hot wind that’s been blown over the last year. For all the jeering of the “Pivot bois” and for all the condescending confidence the interest rate analysis bears have presented - this is the actual facts for the matter.

I can’t help but feel someone who calls this area of analysis their home should have mentioned this. I’ve read analysis of people who call this their niche. They should have mentioned this. In hindsight, this is stupidly obvious and I should have checked it myself but it heavily discredits analysis from people who use this metric and didn’t mention it.

None of this affects my analysis in any way at all. To be honest with you, I don’t even know what the US interest rate is right now. I think it’s 5%. If it is, I know that in the same way you know the words to an advertising jingle. You don’t seek it out, you just take it in because it’s there. If the rate is not 5%, then that just shows how little I care.

Not knowing these things has not inhibited me from having predominantly the correct bias in all big swings over the last couple years and to find out interest rates can be seen as (At the very best) a neutral indicator is not going to change the sequence of rules I am going to use to make my next decisions. I just think you should know.

People speak about interest rates as if it’s a 100% signal and it’s actually a 20% signal in which the max historic drop is 17% and we’ve completed 18%.

It might be different this time - but there’s actually no historical basis to support a short SPX thesis based on interest rates.

We’ll carry on as normal. Same decision tree to determine if we’re bull or bear. I just wanted to let you know that the persistent objections to bull analysis that have been raised since the end of 2022 are actually invalid. The real market reactions do not predict the same as the predictions of those talking about.

Please note: I draw lines on charts. It may be I missing important context here. I don’t know the interest rate today so I’m not exactly schooled up on the 70s (Before I was even born). I am open to learning if people can explain things I do not understand about this.

But people should also know the real facts of history.