Fail to prepare, and you’ll prepare to fail.

Genius is 1% inspiration and 99% perspiration (incidentally, that will bode well for the general geniusness of folks living in London right now).

At this point in the season, the data is pretty thin on the ground. You can take two data points on a chart, and draw any sort of line you like between them, tell any sort of story that suits you.

And so I’ve been resisting the urge to be too declarative yet. I’m mostly just doing the preparation for the season ahead. Laying groundwork. Putting in some framework to consider how the season unfolds when the data is less sparse.

I introduced my Jagalytics Heartbeat as a way of tracking the team’s stats against historic high performers (and as a way of seeing how Wilsonball evolves season-on-season).

Then last week, I introduced my Jagalytics Index to do the same for players. Which players are excelling in particular roles?

This week, I’m going to reflect on one of my models - Biased Outcome Indicators of Non-played Games (or BOING … geddit???). Over the last couple of season, I’ve run this as a way of forecasting how the league would unfold.

One of the most important things you can do with a model is review it. See what it did well, what it failed to capture, and then come up with ways of improving that over time. So in order to prepare for the season (I probably won’t run BOING til everyone’s played five games), I thought I’d reflect on last season.

What it does

BOING is what’s called a ‘Monte Carlo’ simulation. The idea is that the world is so full of uncertainty, and most real-world problems are subject to so many variables that no analytical model can accurately measure what’s likely to happen.

Instead, then, we can set a menu of ‘initial conditions’ - the starting parameters that describe the general shape of the world (roses are red, the Pope’s a catholic, bears are known for arboreal defecation and so on) - then introduce a degree of randomness, and let the world unfold. If you do it once, you’ll get one possible outcome.

If you do it ten times, letting randomness tweak your initial conditions every time, you might get ten different outcomes, or you might get one outcome that happens more often than the others.

If you do it hundreds, thousands, tens of thousands of times…you begin to get a ‘probabilistic picture’ of what’s likely to happen. 23% of the time this happens. 4%, that happens. 49% of the time, the other is where we end up.

It’s like those old kids’ toys with pins on a board. Drop the ball bearing in, and it bounces here and there, left at that pin because a butterfly flapped its wings outside, right at the next because Aunty Ethel sneezed.

At the bottom, you find out which bins are most likely based on how many balls end up in each of them. BOING follows this same approach. We can’t possibly model all the injuries, and weather, and disruption due to a flare on the pitch, or a freak gust of wind at Gayfield, or the ‘keeper having a bad night’s sleep and so letting his concentration slip.

But we can look at form both recently, and over the season, and use that to set some initial conditions. Each team gets given a range of possible points per game, based on historical evidence, and at the start of each simulation, they are given a random number of points per game for the remaining matches of the season within those ranges.

It’s a good way to go because you don’t then have to pick up on every nuance. But a team who’s done well so far is more likely to continue to do well, because you have some evidence that they can get points. A team whose recent form is strong is more likely to sustain high points hauls than a team who’s lost four on the bounce is to turn their form around. It can go the other way. But it’s rarer than good teams staying good and bad teams staying bad.

The randomness means that you allow for that not to happen every so often.

And because all ten teams get the same treatment, each time you launch a simulation, you get a slightly different outcome. A team who’s on a five-game winning run is statistically more likely to get a good rate for the next few games, but even if they get a randomly low number, everyone else will simultaneously have to do well to overtake them.

Then, after doing this 100,000 times, you get a sense of the most likely outcomes. I normally express it in terms of likely outcomes:

• Team X gets at least Y points in half the simulations;

• Team Z finishes in the top 4 in 60% of the simulations.

The range of points lets you see how likely one team is to overtake someone else.

For instance, in early April, here’s what the model thought:

St. Johnstone tended to have the most points in simulations, but the likely range for Thistle and St. Johnstone overlapped quite a lot. By that point, Dunfermline and Arbroath were likely to be third and fourth, with the Pars looking at a higher total in half the simulations, but Arbroath had a much narrower ‘spread’ - the model was more confident in what Arbroath would do, and Dunfermline’s end to the season was likely to be more chaotic.

Basically: BOING is a way of taking some evidence from recent history, introducing randomness, and trying over and over again to see who comes out on top.

It’s about probability. A 30% chance of winning the title means that in 3 out of 10 sims, the team in question outperforms everyone else. Last season’s version (BOING 2.0) didn’t actually simulate games per se. It just considered remaining resources (number of games left) and guessed at how many points per game the team would get.

This is not a digital Paul the Octopus. It’s more like Peter Snow at the start of Election Night, saying “Just for fun, if this result rolled out across the country, here’s what parliament would look like…”

So: how did the model do?

A one-and-three-quarters-horse race at the top

The first thing to notice is just how long Thistle and St. Johnstone were in a title race, as far as BOING was concerned. When I first published a run of the model after five games, St. Johnstone won the league in around 40% of sims, Raith Rovers (remember them?) and Arbroath in just over 20% each, and Thistle in under 10%. That’s what happens when you have a slow start to the season the way we did last year.

But as we moved through the first quarter of games, and then into the second quarter, the two teams grew gradually closer together, and away from the teams behind, until the published model in late November had the teams almost 50-50, each winning about half of the simulations.

As far into the season as game 24- two-thirds of the way through the season - Thistle were still winning more than 36% of the sims. St. Johnstone were normally ahead, but the two teams were mostly sweeping up the title and second place between them in a high number of simulations.

There was some volatility, with St. Johnstone climbing to 70%, back to 57%, and then climbing again.

But from fairly early on, the model settled into seeing the Championship as a two-horse race - two credible contenders, who both had a not-insignificant chance of winning.

By game 28, St. Johnstone won 99% of the sims, and really the season was done for BOING. But even then, Thistle were consistently the second strongest performer, almost always finishing in second.

Interestingly, the model was very strong at forecasting the points the teams would hit. From late October onwards, Thistle’s projected tally was between 63 and 68 - and we ended on 66. St. Johnstone were slightly underestimated, with most of the season projecting them to be just over 70 points, and in reality, ending up on 77.

By mid-March, both teams were projected to get to a count that was within 3 points of their final haul.

All this to say: although there was some volatility in the probabilities, for much of the season, BOING was very strong in having these two as the top two - by quite some distance - and always preferred St. Johnstone. And it was good at forecasting just where they’d finish on points.

Race to the bottom

The second area that BOING did well was in the relegation dogfight.

Early on, BOING was very worried about Airdrie - which isn’t that surprising. By game three, 77% of sims had them in the bottom two by the end of the season, and it was over 97% by game six. It bounced around, and there was a brief rally towards the end of the first half of the season, but then they were around 80% or higher to be in the bottom two for almost the whole of the rest of the season.

Of the other candidates, although QP and Morton both had periods of flirting with the drop, very quickly it became obvious that the other team in real danger were (eventual 10th-placed) Ross County. Even whilst I was still assuming they would turn it around, their probabilities in BOING kept pulling them downwards.

By the midway point, they were in the bottom two in almost 9 out of 10 sims, and after a brief resuscitation when Kettlewell arrived, from game 24 onwards, they were always 70% or more likely to end up 9th or 10th.

A short run of results can make you feel like things are changing - that’s what County experienced when Kettlewell first took over. It was closely packed enough that it felt like they’d pull it around. But in the end, BOING’s instincts were right. Both teams ended up where the model guessed.

Playoff hopefuls

What BOING struggled with a bit more was the playoff battle. Arbroath did fairly well from quite early. From game 9 onwards, they grew from a 40% chance to make the playoffs in a steady upwards trajectory. By game 15, it was around 90%. And from there on, they just locked in the chance that they’d finish in the top 4.

Raith Rovers started off well, but by game 15, they dropped away with their chances of a playoff hovering around 10% for several months, before disappearing (apart from a minor palpitation in the final few games).

Ayr, though, bounced around. They were up near 70% by the end of the first quarter, but by the end of the second, down to 20%. Up again by game 24 to 60%, and then cratering to almost zero by game 28 onwards.

Dunfermline were even more chaotic - every three games seemed to swing from really likely to make the playoffs to ‘not a hope.’

BOING reacts quickly by design. It pays attention to form, and weights its thoughts accordingly. Of all the teams in the league, Dunfermline’s form blew most hot and cold. Ultimately, it was good enough to get them into a very strong place. But BOING couldn’t work them out. It understood Thistle, St. Johnstone, County, Airdrie, and Arbroath pretty quickly. But Dunfermline? Nope.

Timeliness

The big question for any model is not just how good a predictor it is, but how quickly it becomes a good predictor of what’s going to happen.

The season evolution of BOING’s accuracy can be measured in two ways - you can see them both on the chart above. How quickly did it figure out what zone of the table teams would end up in (Champions, top 2, play-offs, relegation threatened, and so on)? And how closely did it forecast a team’s eventual points tally?

BOING had a season of three thirds (no, Mr Infantino, that was not a commercial suggestion. Step away from the rulebook, please).

The first 12 or so games, it was finding its feet. Under 60% of the teams were predicted in the zones they ended up in, and the average team projected points was out by 7 or more points.

After that, up to game 25 or so, 60-80% of the teams were commonly projected to end up where they really did, and the average error in projected points was more like 5 or 6.

From there on, BOING got the final position right for at least 90% (9 of the ten teams) each prediction, and the average points prediction was only off by 2.9 points (down to 2.3 by game 30).

So by ten games out, BOING was very accurate. By midway through the season, it was pretty strong at forecasting where things were going. I’m genuinely happy with that.

What next?

BOING 2.0 seemed to be good at getting the lay of the land - the general shape of the league - very quickly. St. Johnstone and Thistle were the leading pair from early on, and they behaved differently to all the teams below them by the end of the first quarter.

It remained positive about Arbroath, and pessimistic about County and Airdrie, despite fluctuations in form.

But the thing that strikes me is that, for the fine margins, BOING 2.0 is moving around too quickly. It was very uncertain about Dunfermline, and couldn’t quite work out what to do with Raith Rovers, Ayr, and QP. It was a touch too bullish about Thistle’s title challenge at points, and a bit too conservative about St. Johnstone’s points tally.

I think the next stage is to make it a little bit smarter. Not to try to model every possible parameter, lest we lose the areas where it did well. But I think I can make it a bit better in at least three ways:

1. Make it more aware of the underlying performance, and not just the points in recent games. I’ve written in the past about the fact that xGD in the trailing five or six games is a very strong predictor of the upcoming PPG (which is why it’ll be in the Jagalytics Heartbeat too). I’ll probably work that into BOING 3.0.

2. Make it a bit less influenced by recent form - making it smarter at asking whether recent fluctuations are just a quirk, or a sign of something deeper

3. Give it a better sense of what lies ahead - not just simulating points per game, but giving it a feel for upcoming game strength, and making it better at being a ‘zero sum system’ so that it actually takes into account the features of games to be played

I suspect that’ll iron out a few of the rough edges.

These are all sensible ways of making BOING better at simulating the real world.

But then again…when has the Championship ever been sensible?

A final aside

If you haven’t already, can I suggest you go read Aaron Richardson’s page? He’s written his dissertation and put much of the content online.

Among other things, he’s interviewed players and others around the club about the way they engage with the community Thistle exists in - from Alex Samuel on being a Christian at a club in a city that often handles football and religion pretty poorly1 to Lee Ashcroft thinking about responsibilities towards fellow pros, and other great articles covering an Iranian in Glagow, Jags for Good, and the longest serving manager in the SFL, there are some great reads. It’s really precious to have such access to the ‘real people’ involved in our club, and I’m grateful that Aaron not only investigated football and community, but that he’s shared what he’s discovered. A great read.

Or, in football parlance, the returning player is ‘better than a new signing’.

Recently, I posted some numbers quantifying Ben Stanway’s season last year. No central or defensive midfielder in the last 3 seasons of Premiership, League 1 or League 2, or the last 8 seasons of the Scottish Championship, managed to put up numbers comparable to Stanway across both defensive disruption and attacking threat. Several players came close on one or the other, but no-one managed both.

He’s a very rare talent, and one we’re unbelievably lucky to get to watch for another season. His team-mates clearly love having him around, and the whole club has a buzz to have him back. It’s like, as the man said, having a new signing.

Having Ben back made me think about indices (indexes?). Normally, when you think of an index, you think of the back of a book - a list of topics and names and which pages they appear in. A lot of software uses an index too - a way of looking up data more efficiently to reduce the computing power required to find what you’re looking for.

I used a simple type of index back in January in my red and yellow lines series (e.g. here) - using a set of parameters to filter a pool of players to those who are relevant for a particular type of role.

People really responded to that post, and I like the general idea (some folks on Football Twitter, like Spencer Mossman, do a great job of proposing new indices). I’ve decided I’m going to add to my Jagalytics Heartbeat concept as a way of tracking the season by introducing The Jagalytics Index - a player-level way of tracking players who are having a special season.

What’s in an index?

Each index will have the same set of criteria:

1. It must be based on a Thistle player - current or historic. I’ve taken the last few seasons of data.

2. The player should be compared to a cohort of players who play in a similar role, in the Scottish leagues, with a reasonable number of minutes played in a single season to qualify. Each player might appear several times in the cohort, if they’ve hit the minutes threshold in more than one season.

3. The index should be based on only 3-6 metrics maximum. That’s about the right number to capture what the combination ‘means’ in the real world.

4. The combination of metrics should be something that the player excels in (or did in at least one season)

5. The combination of metrics should be something that distinguishes the player well (if you filter the players in the cohort to at least as good as the Thistle reference player in the metrics, you are only left with at most 5 or 6 other players). It’s not enough to be good at something if everyone else in your position is too.

This leaves us with an ‘index’ - a way of filtering the current season of players for someone who matches up well (i.e. they hit all of the same numbers the reference player did), as well as giving them an ‘index score’ - out of 100, how impressive is this player’s season compared to the relevant metrics.

Let’s have a look at what I’ve chosen.

The Stanway Index

Reference player: Ben Stanway, 25/26 season.

Role cohort: CMs and defensive CMs (based on the primary position in Wyscout data)

Metrics:

• Shots per 90 minutes played

• Touches in the opposition box per 90

• Successful defensive actions per 90

• Possession-adjusted interceptions p90

What the index means: This combination of metrics is unusual for a deeper-lying midfielder. To be high on all four (even just the first three, dropping the final metric, actually), you need to be able to power forward regularly, get a lot of shots, but deliver strong disruptive output too. Ben’s season last year for the first three combined isn’t matched anywhere in the data. By anyone. At all.

It’s way too early to be looking at players who match up well - I’m just laying the groundwork now for later in the season. The idea is that this index helps us find players who are showing the same sort of attributes as Ben did last year; a quick-look to find those two-way midfielders. Normally, I wouldn’t pay much attention to the numbers until the minutes hit at least 600 (and preferably 900) minutes played.

The Tiffoney Index

Reference player: Scott Tiffoney, 22/23 season

Role cohort: LW and LWF to build the index, bringing in the right side equivalents afterwards.

Metrics:

• Dribbles p90 (which is the obvious feature of Tiff’s play that season)

• Progressive runs p90 - these are runs that move the ball significantly towards the opposition goal. So not just short burst runs, but runs that are designed to drive towards the goal. These sort of runs need to drive forward at least 10m inside the opposition half.

• Touches in the opposition box p90

• Expected assists p90

• Expected goals p90

What the index means: This is a running attacker who provides end product from out wide. He takes on his man again and again, but does it in a way that forces the ball towards the goal, not just showcasing fancy footwork. Crucially, the player gets into the box, and has end product - the first three conditions are met by a few players, but very few also manage to hit a high rate of creating chances (xA) or high quality shots (xG) too. This index looks for players who do the first three, and then one of the final two. Tiff was very unusual in that season, because he did all five. We’ll end up with ‘creating Tiffs’ and ‘shooting Tiffs’.

I’ve split the index to look for primarily left-sided players, and primarily right-sided ones.

Left

Tonight’s opponents Livi have a player who is already red on all the numbers (though of course that’s based on a single game, so … take it with a pinch of salt). Recent Morton release Owen Moffat (who I’m a fan of) also scores well. And Fitzy is in the top 4 from the first weekend. Historic reference points outside of Tiff are limited to two Celtic players.

Right

The right-sided flavour contains only Celtic players again (in the historic data). Keep an eye on Elgin’s MacIver this season to see if he keeps up those numbers.

The Harry Milne Index

Reference player: Harry Milne, 22/23 Championship season

Role cohort: LBs to build the index. LBs and RBs to find players subsequently.

Metrics:

• Progressive runs p90

• Ground duels won p90

• Dribbles attempted p90

• xA p90

What the index means: This is a battling fullback, who gets stuck in, but who runs the ball exceedingly well, and generates a high level of chance for his teammates at the end of the run. Not just a ‘bomb forward at all costs’ guy, but one who battles for the ball out of possession too. Not just a ‘headless chicken’ runner, but one who covers a lot of ground, and crucially gets a pass that creates a great chance.

Left

It’s very hard to find players who hit the Milne Index. In fact, only Milne himself at Cove and then again at Thistle hit these numbers historically. And Milne himself pops up with his first game for Hearts this season. It’s noticeable that the players on the top 10 so far are almost all upper Championship or Prem players.

Right

Again, the historic index picks out only a single Celtic player. Interestingly, whilst several left-sided players this season started well on the chance creation metric, it’s more the progressive running that pops on the right side. There are a few lower league players in this set, which is worth keeping an eye on.

The Ashcroft Index

Reference player: Lee Ashcroft, 25/26

Role cohort: Primarily CBs

Metrics:

• Defensive duel win rate

• Aerial duels won p90

• Interceptions (possession adjusted to allow for lower opportunity if the player plays for a high-possession side)

• Progressive passes p90 (equivalent to the run metrics above, but focused instead on passing accurately in a way that successfully moves the ball 30m+ down the pitch inside your own half, 20m+ across the halfway line, or 10m+ in the opposition half). This is often correlated with the high-performing Championship teams.

What the index means: This is a combative, engaged defender who handles business out of possession, reads the game well (snuffing out passes), and is crucial in build-up play.

Our captain had a very unusual season last year. To be as dominant out of possession, as involved in early buildup, and maintain his historic ability in intercepting passes, is very hard to keep up over a season. I’d expect this to drop off significantly as the season goes on, with very few players being able to keep this up week in and week out.

The BBG Index

Reference player: Brian Graham in our League 1 winning season (20/21)

Role cohort: Players who are primarily a CF

Metrics:

• Goals p90

• xG per shot

• Aerial duels won p90

• Touches in the box p90

What the index means: A high end-product forward, who gets consistently high-quality chances, and has aerial presence.

Interestingly, BBG himself appears at the top of the list after game 1 - he’s still doing Brian Graham things up in Dingwall. This isn’t an all-round striker, but a player who will … y’know … score your goals. It’s unlikely many players will cross the threshold - more than a goal every other game is … tough to deliver over a season. So I’m mostly interested in players who get close, and who do well on the other metrics.

The bonus ‘all round threat’ Index (name tbc)

I first started playing with this sort of idea after seeing this particular index a couple of years ago. Those who’ve read Jagalytics for a while will know I’m a big fan of ScoutedFtbl, and how they think about the game. They blend the eye test - how football really looks in real life - with data and analysis in the right sort of blend.

This particular index is a reliable way of spotting players who have hit a combination of all-round threat by midway through a season. Last year, very few players in the Championship got close (though our own Tony Watt was one of them). The original post highlighted several high-end players - Anderson, Semenyo and so on all commanding huge fees in the last couple of years.

So I thought I’d add this into the mix to track as the season goes on. I’ll start with the 4-metric 20+ threshold, and then move onto the 30+ number for Dribbles, Aerial duels, and Shots (which is a good second level filter for the real elite performances).

As Jake Entwistle said in reply to my tweet, it’s a reliable way of finding players who are all-round threats. It’s normally too late to help recruitment (by the time players put up these numbers, everyone and their dog know the guy is a pler), but it’s a helpful barometer. So far, there are 17 players tracking to break the 20+ limit in all four if they keep it up and play lots of minutes, including our own Fitzy.

Tracking the season ahead

So now, I have two different ways to track the season ahead:

• At a team level, the Jagalytics Heartbeat

• At a player level, the Jagalytics Index

I plan to come back to these in a few weeks to refresh how things are going. But for now, let’s focus on an exciting Friday evening under the lights, and a very dangerous Livi team adjusting to life in the Championship. Let’s hope Ben can show us all that he’s better than a new signing.

We’ve reached that most magical time of the year. A moment ‘twixt1 Before and After. A sliver when anything is possible, dreams are yet to be crushed, and thoughts that ‘this might finally be the year’ gain momentum.

Tomorrow sees the start of a new Champ season, and so - once again - it’s time to think about football.

Today, I simply want to introduce an idea I’m going to try to track and develop as the season goes on. I’ll add more detail and dig into the thought process behind it (and probably tweak and change it as we go), but for now, here’s the thing… introducing:

The idea is simple. I’m going to pick a (relatively) small smattering of metrics that we’ll trace through the season, describing something of how Thistle are playing. Up and down, side by side, we’ll get a sense of how we’re performing against a tangible reference point.

Each of them will have a comparison to last season (so we have a numbers-y way of comparing Thistle 26/27 to Thistle 25/26), and also a benchmark: over the last few seasons, what would you have seen from the team that won the league, the teams that were in the play-offs, the teams that finished outside the playoffs but not in relegation trouble, and the teams who were in a relegation battle?

Have Thistle measurably changed compared to last season?

Are Thistle playing like champions or like playoff contenders? Or worse?

I’ll track it semi-regularly, and use this as a jumping-off point to consider strengths, weaknesses, and interesting evolutions.

I’ve got six different areas it’ll cover:

Outcomes

This is the main thing. Only one metric here: points per game. Over the last three years, you need to hit roughly 2.0 or more points per game to win the league. You can do it with lower numbers, but there’s a pretty big gap between the normal range to guarantee playoffs, and the range that ensures you’ll win the league. Anywhere in between is … risky.

Last year, we were between those two ranges with 1.83 PPG. But we need to turn that up a notch to be in with a chance.

Underlying numbers

Again, this has just one metric: expected goal difference (or xGD). I wrote about this a bit in January - you can see a link to the post below. Basically, though, if you look at the recent trend of a team’s league xGD (the expected goals they generate minus the xG they concede to opponents), you have a pretty good idea of what their upcoming points per game will be.

If you hit over +0.3 or +0.4 over a few games, you can expect to average 1.8 PPG in the next few (ish). When I wrote that article, we were fresh from a fairly weak league run (drawing with Raith Rovers and QP, just squeaking past Morton, losing to County) in the league, and a loss to Caley Thistle in the Challenge Cup, but the underlying numbers suggested we were in line to hit 1.8 PPG in the upcoming games. In fact, we won two cup games and averaged 2.3 PPG in the league.

That case is anecdotal, but outside this one example, the evidence is actually quite strong on this. I’m working on improving that model a bit, but for now, the heartbeat will include the xGD (possibly the ‘recent trailing’ games, but I haven’t decided yet).

Last season, we averaged +0.35 xGD - again ahead of normal playoff performance, but well short of the normal Champions range of +0.89-1.21. That’s the target for us this season.

Attack

For attack, I’m starting with three metrics:

• Shots per 100 box touches. This is a bit “lies, damn’d lies, and statistics” … it’s a measure of how much a team has to footer around to find a shot. It can be low because the team snatches at the first shot they find in the box, or because they are incredibly decisive. It can be high because the team is carefully looking for the right chance, or just incapable of translating ‘being in the box’ into ‘posing a threat’ In other words, it can be good, or bad. Champions have a fairly wide range here, so this is more a way of tracking how different (or similar) we are to previous seasons. Mostly, playoff teams are slightly more decisive, needing fewer touches to get a shot than mid-table teams, and it could be argued that Champions were a bit more decisive again, but it’s not as clear-cut as some other metrics. Thistle were around 65 shots per 100 box touches last season - a bit lower than the non-champion average.

• Box-touch share. This is a bit like ‘pitch tilt’ - it’s a measure of what percentage of touches in the box go to one team over the other. A higher percentage means you get more of the attack in games. Champions are typically miles ahead here - 63-83%. Thistle were down at 55% last season, so we need to shift that a bit. This is a good way to tell if we’re reallly serious about challenging for the title.

• Spreading the shooting load. The idea here is to see what fraction of the xG falls to one particular player (the most shooty player in the team). Often, the champions are better at spreading it around, with several threats on the team. Playoff teams often ride on the coat-tails of one star forward who takes most of the shots (think BBG). Thistle were at 17% last season. We’re looking for whether the team spread the load around, or focus on one player.

Defence

In defence, again there are three metrics:

• xG allowed per 100 final-third passes by the opposition. Effectively, this is a measure of how quickly the other team can go from breaking through your defensive midfield to getting quality shots. A higher number means the other team don’t need many passes to find good chances, and can go pretty directly through your back line. It’s ok to sit high defensively if you can track back, or your keeper can sweep well. Generally, the lower this number is the better. Thistle had a relatively high (i.e. bad) number for this last year - more like a mid-table side, which is interesting, given the strength of our defenders. We need to figure out how to make it harder for the other teams to go from passing our high line to getting a clear shot.

• OPpDA (or opposition passes per defensive action). This is measuring how eagerly you harass defenders - a lower number means your ‘press is more intense. It’s measuring how many passes the other team gets before you put in a tackle, or contest the ball, or block a shot, or intercept a pass, or clear the lines. Generally, champions have a lower number here - they get the ball back quicker, stop opponents playing their game forcefully (interestingly, Falkirk had a bit of variance home and away on this, but that’s for another day). It’s easier to do if you have a fit team, who’s had a full pre-season, and have depth you can swap in, and it’s hard to sustain if you have an older or smaller squad. Two years ago, Livi used their artificial pitch to help them here. Thistle were less intense than you’d expect last season. Again, this is not necessarily ‘more intense is good’ - it’s more contextual to see what’s changed.

• Average defensive line. Again, this is more to understand how we’re playing rather than predicting the chance of winning. Normally, the champions are able to play a much higher defensive line (normally 33-35m) than anyone else, partly because they’re attacking more and have more of the ball in more advanced locations. Everyone else in the league is normally 30-32m. Thistle were around 33m last season, so closer to the lower end of how the champions would play, but not quite making that final step.

General possession

There are two general possession numbers to watch:

• Average possession (which is pretty obvious). Generally, more of the ball correlates with more success. We averaged around 53.7% last season. Good, but not elite.

• Progressive actions per 90. This is a measure of how many times a team moves the ball significantly up the pitch successfully per game by passing or running - 30m or more up the pitch by the end of the action in the defensive half, 20m or more across the half-way line, or 10m or more in the attacking half. I’ve written about this too, but generally the team that passes and runs the ball most effectively and often up the pitch wins the league. Thistle were squarely in the champions range for this last season with 93.45 passes and runs per 90 that met this target. We need to keep this up.

Set pieces

Finally, I’m going to track set pieces. This is an area where we’ve not been as strong in recent seasons, and where we probably need to improve a lot this year to make that final step up the league.

I’ve got three metrics, all expressed as a ‘balance’. The idea for each is to compare how good you are, minus how good you let your opponents be. A good set-piece team knows both how to create and capitalise on chances, but also how to restrict the opposition. So a positive number here means you’re better at this aspect than you allow your opponents to be. Negative means you allow your opponents to be better.

• Opportunity. How good are you at creating set piece chances (and restricting your opponents from getting them)? Corners, throw-ins, free-kicks. Each set piece is a chance to create a shot or a goal. Champions tend to do well here, and everyone else is some variation on breaking even or slightly negative. The message is: if you know how to create and restrict set pieces, you’ll win the league. Thistle were quite good at this last year - better than a normal playoff contender, but still well short of normal champions. So we need to keep up our progress, and improve where we can.

• Efficiency. How good are you at turning a set piece into a shot? There’s more variance here in the champions, but generally, higher is better. Thistle were roughly like a high mid-table team here. We need to be more efficient at creating shots, and tighter at restricting opponents, to be in contention.

• Effectiveness. This is a measure of how well shots on target from set pieces are scored. Thistle were … bad at this last year. Like spectacularly bad. If we can do one thing to help ourselves this year, it’s improve our goalscoring rate (and lower our conceded rate) from shots starting from set pieces.

Putting it all together

So that's our heartbeat metrics.

I’ll use them to track Thistle’s evolution this season, and open a window onto how our season is unfolding. I’ll return to this, and may tweak and change the numbers we look at as we go.

For now, though, savour the sunrise of a new league season. Taste the nectar of possible promotion, free of the bitterness of unfulfilled hopes. It’s a new day.

Be still, my beating heart.

As I write, Thistle have just won a penalty shoot-out (gasp). No, really. Drama, thou art found at the 12-yard spot.

Football is a low-scoring, high-stakes sport. It specialises in drama in the same way that Breaking Bad does. Sometimes sudden and shocking, sometimes as the result of sustained and building tension. The thrill of a last minute equaliser, or a breathtaking comeback. Sneddon scoring five minutes into injury time. Ethan Ingram clearing off the line.

The drama of football is one of the reasons we love it.

And yet it’s sometimes hard to articulate to someone after the game where that drama came from. It’s not from a single moment, or an isolated event. Not really. It’s a confluence of activity; ebbs and flows, a complex narrative in whose flow sit the weighty rocks of the memorable moments.

For instance, there was a great animation someone created this week about how the England vs. Argentina World Cup semi-final this week changed after England took the lead. You can see it here. It brings to life what you saw with your eyes if you watched the game. A team on the front foot went ahead, and then surrendered territory almost instantaneously.

I’m a sucker for a novel way of turning data into a story. NASA have ‘sonified’ some of their telescopic data (partly to allow visually impared communities to engage with the beauty of their imagery). Or Charlie Clark, who championed the idea of taking the average colour of each frame of a movie and turning it into a ‘barcode’ at The Colors of Motion. From the heathery pale purples of The Good, the Bad, and the Ugly to the distinctive Anderson bold palette of reds and browns and yellows in The Grand Budapest Hotel.

I love the creativity of taking some data and using it to give a fresh look at something you’re familiar with.

I came across a wonderful project this week by Zeh Fernandes, doing just that. Zeh wanted to represent the drama of World Cup games in visual representations of the data - giving a sense impression in a glance of what you would have absorbed if you’d watched the whole game.

I thought this was truly gorgeous.

I found myself going back to the posters again and again, captivated by them. Even games I hadn’t watched caught my eye, and made me dwell for a moment on the drama of an event I didn’t participate in.

So I though I’d give it a go.

The method

I’m no graphic designer. I once attended a great seminar by the Guardian data journalist team, and one of the points they made is that good data storytelling involves at least four steps - data, story, chart, and design. And, the seminar leader said, almost no-one is good at all four steps - at least, not at the same time, on the same project. It takes a village to raise a child, and it takes a team to tell a story with data.

On my best days, design is my weakest of the four disciplines.

So, confident in my ineptitude in terms of design, I decided to follow Zeh’s methodology, with a few tweaks of my own. Any strength in the visuals you see here is tribute to Zeh. Any failing is down to me.

I used an able agentic deputy of Mr C. Odex1 to help me build the code and iterate (because it allowed me to work on it whilst standing on the sidelines of my youngest son’s football coaching session).

The basic idea is this:

• Much of the drama of a game of football is in the ebb and flow of attack: the rearguard defensive action, or the tsunami of final third threat. So let’s visualise the match through the lens of attacks (and especially shots).

• Turn the timeline of the match into a grid rather than one long line. Zeh played around and found a grid that goes from top left across a row, then to another row, and another, and so on, worked well. In the end, he had an 8x11 grid, so each ‘box’ represented just over a minute of the game. I stuck with that.

• The main currency of attacking drama was shots - and so each shot was dropped into its time-coded box

• Each shot then had a set of lines streaming from the side to the shot-dot (home side from the left, away from the right). That helped give a sense of how competitive and confrontational a game was, because it feels more like two armies attacking each other

• The lines were longer if the team had a higher share of the game’s passes, and thicker if the team had more possession

• Additional ‘secondary’ events were added to the grid when fouls and so on happened

• When teams scored, those dots then exterted ‘gravity’ because goals are the most influential part of a game. More dramatic goals - late winners, climaxes of a series of pressure, big counter-attacks breaking out from sustained opposition pressure - impacted the chart more, and affected the chaos and busy-ness of the streams.

• This gravity caused the lines streaming through the page to ‘bend’ like a planet on its moon or a black hole on passing rays of light. Home team goals ended up causing lines to swirl in one direction, and away teams the other, causing a bit more ‘entropy’ to the plot.

Zeh used this process to create the basic ‘pattern’ of a game, and then blends the streams into colour-smeared posters. They really are beautiful.

I worked on a similar logic, with a few changes.

• I used Championship data from Scotland, and particularly Thistle’s games from the season just gone. My aim was to build a poster for each game I had data for (34 of the 36 Champ games from 25/26)

• I didn’t use secondary foul data, but I added in substitutions, as those often change the flow of the game

• I weighted the shots based on non-penalty post-shot xG (my own model), which impacted the density of the the lines streaming to the dots. That meant that big chances create more visual impact.

• I gave penalties their own icon, and missed penalties or missed chances with high xG, got ‘anti gravity’ - they pushed the lines away from themselves, rather than towards them the way goals did. This added another element of entropy to the distortion of the lines.

• I didn’t end up fully blending the lines, as I liked the aesthetic with discernable back-and-forth that the patterns gave

• Thistle always appear as red, and the opposition normally appear as their dominant non-white colour (e.g. Dunfermline are black), unless that was too close to Thistle’s red. In those cases, the team was given a colour that came from last season’s away or third strips, or their logo, that gave a reasonable visual contrast (for instance, white doesn’t work on these charts, and so the natural choice for Airdrie would be red, but instead I picked a purply blue from one of their other strips). The aim was contrast and ‘battle’ drama, rather than game accuracy.

You can see one example above, from Thistle’s win against Dunfermline in April. It immediately caputred for me the strength of Thistle’s start to the game, with limited pushback from Dunfermline’s black, and the impact of two quick goals early in the game. Despite phases of pressure in the second half, the lower part of the game is much more reserved.

If you put the four posters for Thistle’s league games against St. Johnstone together, you can see the different flow of our games. The onslaught of that opening day humbling, the cautious but canny 2-1 win, the more open blow-for-blow 2-2, and a cagey 0-0 in March.

Our 5-0 victory over Queen’s Park gives a different visual impact. The first half started well, but with limited action between about 25 minutes to the break. But the second half … a tidal wave of red. Irresistable and destructive.

For drama, not many games beat the late, late, late Ablade goal against Falkirk in the penultimate game of the 24/25 season. And that too is visible in the poster.

Falkirk’s blue dominates most of the game, other than BG’s early goal. But the visual is sucked towards the drama of that last-ditch winner. Streaking through the ocean of blue is a stain of deep red. The gravitational pull of that winner bends the rest of the game’s narrative.

The game right at the end of that season - just as important in getting us to the playoffs that year - was a different story. A rearguard action, trying not to be overwhelmed by a surge of gold throughout the whole game, intense and constant. As an aside, this was a good example of having to deal with a particular quirk of the data - own goals - and making sure they were captured in the right way.

I loved this idea by Fernandes. It gave me a little side project to plug away at, trying to produce something that was visually appealing. I have posters for each of Thistle’s games where the data was available, and I’ll probably go back to them in future to help me tell a story.

I might share some of them over the coming months, or I may not.

Sometimes, art is enough in itself, and you don’t need a ‘story.’ And data can generate wonderful art, if you know how to look at it.

I often feel a great satisfaction in corraling data into shape to allow me to tell a story. You often don’t see that, but what I imagine a painter feels when her canvas dries is what I feel after wrangling messy data into order. Whilst I don’t always do it justice, it feels beautiful to me in its own way when chaotic and disparate numbers can be organised and structured to help illumniate a narrative. But for once, this time, it was fun just to try to focus on the art of the drama as an end in itself.

The aim was to build each team roughly in Thistle’s playing shape from the 25/26 season (so that in theory you could put them on the pitch and they’d form a coherent side). Each team would have its own defining characteristic: a single criterion that judged who was best suited for the team, as if there was only one recruiting paradigm worth using.

In the end, that means that certain players are less likely to make it in. Players who play in unusual systems, or at least systems that use different shapes than Thistle’s are slightly biased against - and that means that it was harder for players like Andrew Tod, Dylan Easton, and so on to make it onto the list. I’m fully aware that this is the case, but the point of this exercise isn’t to replicate all the other ‘teams of the season’ (otherwise those sorts of players would be very likely to be in contention). I wanted to show how you could build a team using a somewhat coherent - albeit woefully myopic - single datapoint strategy.

I’ve said before that in reality, this isn’t really how data input works for a football club. Real selection and recruitment decisions are made using a variety of evidence, layered and woven together based on the preferences of the football department. Data is one strand, and good data analytics helps you make better decisions not by being ‘clever’ but by helping provide understandable, accessible evidence that real humans can engage with.

Almost no club in the world, I suspect, truly runs on extreme Moneyball rules. Some will lean more heavily into data, especially to find undervalued players or to reduce the inevitable risk of signing a player. I can’t imagine any player is signed because ‘the model said they were 19 out of 20’ … but it’s probably the case that it can feature in the decision-making, alongside the instinct, knowledge, and experience of real scouts, coaches, and industry experts.

This is a simplified form of feeding data into a decision-making process. In my experience, more than half of the work of using data well is wrangling - getting data into the right shape, handling the edge cases and mitigating the risks that you’ve misinterpreted, abused, or unduly warped the data. Another 30% or more is thinking carefully about how to land that data in a way that real people can absorb what it means and helping them to use it to make decisions. Very little is actually doing super clever modelling stuff.

All that to say - I’ve enjoyed this sandbox way of constructing a team based on a limited scope of data. It’s interesting, because it shows - in a vastly over-simplified way - how you can use data to judge between one player or another. But it’s also shown me again the importance of having a target - a system that you want to play, a style, and priority. That allows you to make the messier decisions about which players are truly in and out of scope, and the order you need to make decisions.

I switched to picking wingers before 8s because it made sense to get the wingers in their true positions even if they’d be near the top of the advanced midfielder rankings too. I filtered out some players who were good and viable candidates for the central midfield roles because they weren’t really the style of players I expected. I left out some very talented players in the defence (Chilikoa-Mullen, for instance) because he scored highest in a position where I don’t think you’d actually get the best out of him.

You can only make those decisions if you know what you’re aiming for.

That’s been one of the real strengths of Mark Wilson’s Jags this season - he’s had a real clarity of how he wants to play, and he and the team around him have both recruited to that plan, and coached players specifically to that plan. Good strategy is, after all, what you say no to.

All that brings us back to the final part of our Teams of the Season: the front man. Over the year, Thistle have played with one or two up top, but for the sake of this exercise, I’m going to pick one striker for each of the three teams. In two cases, the player isn’t a real out-and-out centre forward - but I think all three are arguable.

Impact

The first team to be completed is the Impact team. And the central forward for that team is Jamie Gullan. Four players broke 14 on the impact score, and another 4 (including Alex Samuel) were between 13 and 14 - players who made their teams’ outcomes noticeably better when they are on the park. Several of the players in the list have played at CF over the season, but aren’t normally a forward (Chalmers and Fitzpatrick both feature on the chart, for instance). McPake again scores highly here, but as we already have him in the side, he’s not going to ‘win’.

Gullan isn’t really an out-and-out striker. He excelled in his final year at Raith surrounded by players like Easton and Finlay Pollock, providing a big creative foil for the goalscorers around him. But whilst others at St. Johnstone have gained plaudits - McPake for his goals, McAlear and Holt for running the midfield, Sidibeh up front before he left, I tend to think the most important signing that St. Johnstone made last summer was Gullan.

He is exceptional in buildup play, linking and connecting St. Johnstone’s attack. 14% of his passes are ‘progressive’ - moving the ball significantly closer to the opposition goal. He generates around 0.73xA per 100 passes - meaning that he only needs about 135 passes to create the quality of chances that lead to at least one expected goal for his team-mates (very few players in the league are above 0.6). He delivers crosses when needed, and wins around 60% of his dribble attempts. He averaged an assist every 4 games, and around 4 goals every 10 games.

For St. Johnstone, he didn’t have his best games when he was the main striker, but by this measure - a player who has played as a striker at some point - he is well ahead of the other options.

Undervalued

Moving on, then to the undervalued player, there are again a cluster who are in contention. Alex Samuel, Jamie Murphy, Mark McKenzie, Nikolai Todorov, Tomi Adeloye, and Gullan all broke 15 for how much impact they delivered given their market value.

But topping them all was Euan Henderson of Airdrie (well … not any more!). Henderson’s impact wasn’t necessarily that high. He played for a team that struggled, and under several different managers. Sometimes he was on the pitch for the harder moments, and so the impact score doesn’t look as compelling. Again, he was more of a wide forward and occasionally a central striker than a true CF.

But he was supremely capable at finding shots, and a very dangerous forward nonetheless.

As I’ve already said, I’m willing to put guys in who aren’t always used in this sort of role. Tony Watt was used as a 10 and a left winger at times this season - in much the same way as Henderson often is. But he is also a very effective close-range striker. The same could be said of Henderson.

Henderson was willing to track back more than most forwards in the data, and won his defensive duels well. He’s not a typical striker, but he is very effective at drawing free kicks from opponents, and excellent at generating shooting opportunities by how he carries the ball. He’s excellent at finding space to receive long passes too, which helps him quickly find gaps to shoot.

Wildcard

The final team to be completed, then, is the wildcard group. For this team, there isn’t a single coherent metric across all positions. Rather, I picked a role-appropriate metric for each position. This time, it’s all about goals. I built a metric (as I’ve done before) that’s about being clinical. It is mostly focused on how many shots are needed to get a goal, with some flavour of average shot quality, volume of shots, and how many touches in the box they need to find a shot.

Above zero means the striker is more clinical than average. Above 1 means he’s more clinical than about 67% of strikers.

The winner of this one is Josh Fowler - probably the most traditional forward in the teams. Although he’s tagged at St. Johnstone here, it takes into account his time at QP as well.

Fowler, on the whole, hasn’t been a high-volume shooter this year, and he often takes quite a lot of touches. But he does that to find a really good quality shot. Not that many forwards in the league this year focused on getting high xG chances. Last year, players like Brian Graham, Ben Wilson, George Oakley, and Anton Dowds all thrived on 0.2+ average xG shots. This year, Fowler and Michael Ruth were the only forwards to focus on that sort of play.

Fowler is a strong technical forward. He gets more lift on his shots than almost anyone else in the league, putting the ball at a difficult height to save. More than half of his shots are on target, and almost a third of shots on target are scored. That’s pretty impressive.

The Final Lineups

That, then, completes our three teams. You can see the final lineups below. Thanks for following along over the last week and a bit - I’ve really enjoyed reviewing the 25/26 Championship season through these lenses. Hopefully, you’ve learned something here and there, and haven’t found the outcome too offensively off-the-wall! Which one would you back in a head-to-head?

I’ll probably go on to write some ‘What to Expect When You’re Recruiting’ posts soon. But for now - well done to everyone who made it into the Teams of the Season.

Impact team

Undervalued team

Wildcard team

Another of my intentions was that I’d focus on each position, gradually working from back to front, with wingers being the penultimate post, followed by the forwards. But as I said in my first of two posts on the wingers, the interchangeability of the attacking units means that it made sense to pick the more rarefied wide players first, and then come back to the more advanced midfielders.

A third intention was that I’d stick rigidly to the style of players in the standard shape of Wilsonball 1.0 - a sort of 4-1-2-2-1 with wingers and two advanced running 8s. In reality, as I dug into the data, that felt a bit restrictive on the type of 8s we’d get in the teams. So I’ve tweaked the criteria a bit to give me two more central midfielders who can either dominate the centre of the park or play a more running 8 more akin to how Thistle use these players. After all, we’re not going to play these sides - we just want a shape that has three central-ish midfielders.

So: best laid plans, mice, men, that sort of thing.

Now we have our wingers, I’m back to the central midfield. Let’s look for two midfielders who are what we can call ‘pushers’ for each of the teams; midfielders who drive teams forward, fend off the early buildup of the opposition’s defence and so on.

You can push by creating threat yourself, or by fending off the opposition at the earliest opportunity, making the central channels uncomfortable.

Whereas the anchor is focused on guarding the defence, and controlling the earlier phase of play, these guys are more about supporting the attackers, protecting the wingers, pushing the other team backwards. They won’t be the standard Wilson 8s (necessarily). I’ve pulled anyone with 1000+ mins and some time in CM or attacking midfield for our pool, which gives us just over 40 players.

Impact

First, then, are our impact players: the ones whose presence lifts the team, and whose production is better than replacement ‘normal’ players in the same league and role.

The most impactful player in the set is Gilmour at Dunfermline - by quite a long way - but as he is already in one of our teams (along with Matty Todd), we’ll leave him to one side. So the next two players are Thistle’s own Ben Stanway, and Harry Cochrane of Arbroath.

Cochrane is an interesting player. He’s not that involved in creating (he only contributes about 6% of the early buildup and 6% of the late buildup), but he’s the most active team contributor on the other side of the ball. He generates more defensive output than any other Arbroath player other than Wilkie and O’Brien. He’s a true pusher in the sense that he forces the other team to work incredibly hard to get through him.

Stanway, of course, has had a breakout season. His shooting is technically strong - improving the xG post-shot by around 8%.

He gets into the left half-space - the channel between the middle of the pitch and the left flank - more than most other central midfielders (about 28% of his events happen here). He reads the game incredibly well - getting more than 7.4 interceptions per game (adjusted for possession) - and he gets into more than 16 defensive duels per game, slowing the opposition down through the middle of the park.

Cochrane and Stanway provide an excellent impact pairing in the centre of midfield.

Undervalued

The undervalued pair are a bit left-field - they’re not who I’d naturally expect to come out of this list. The two added to this team are Ryan Dow and El Banzo. Dow has had a strong creative season for Arbroath. His speciality was pushing through the defence to the byline and cutting it back across the goal for a player to run on to. Earlier in the season, he also worked well taking set pieces.

Banzo, meanwhile, had a solid season at Somerset Park, doing Banzo things. Metronomic passing, bite and aggression, and making life difficult for opponents. Surrounded by several young midfielders, his impact was undervalued according to the model, raising Ayr’s floor in what was a difficult season for the team with reliability and control of the ball.

Wildcard

The wildcard metric I’m going to use for this role is what I’m going to call ‘Positional Attacking Contribution’ (I need a better name for this).

I wrote a bit about this a few weeks back.

The most successful teams have multiple players who contribute to each of two different types of attacking - either ‘in front of the defence’ or ‘behind the defence’. To capture this, I’ve looked at how much time each player spends inside the opposition defensive unit (in the roughly drawn white shape in the diagram below), and how much behind the deepest-lying defender’s average position (the red box below).

I’ve combined these into a single metric and ranked the players based on how above average they are at contributing to either of these two roles.

That gives us our two wildcard midfielders: Seb Drozd and Cameron Blues.

Drozd is a bit different to the others we’ve looked at. He can play as a striker, or a deeper attacking midfielder, or out wide if the need arises. He’s young, but he’s an exciting prospect.

He covers huge swathes of the pitch, on both sides, and is effective at getting into the dangerous areas of the pitch. He gets behind enemy lines in just under 50% of his events, and inside the opposition defensive unit for another 37%. He wins more defensive duels than attacking duels, passes well in the final third, and completes more than 55% of his dribbles. He’s a decent striker around the six yard box, and provides a good, varied positional threat.

Blues on the other hand, contributes lots of interceptions, doesn’t run or dribble as much, but generates a lot of defensive activity. He passes well, and converts around 14% of his shots on target. He’s not the most eye-catching of Morton’s players, and yet he comes up well in this particular metric, contributing effectively in the zones that make life hard for opponents.

The State of Play

Our teams are almost done - only one more player to go now! Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left), Ben McPherson (right)

• Anchor midfielder: Charlie Gilmour

• Pushing midfielders: Harry Cochrane, Ben Stanway

• Wingers: Josh McPake (left), Logan Chalmers (right)

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left), Craig Watson (right)

• Anchor midfielder: Oisin Smyth

• Pushing midfielders: Stuart Bannigan, Ryan Dow

• Wingers: Michael Garrity (left), Matty Todd (right)

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left), Shea Kearney (right)

• Anchor midfielder: Jason Holt

• Pushing midfielders: Seb Drozd, Cameron Blues

• Wingers: Aidan Fitzpatrick (left), Taylor Steven (right)

People object to Players’ Player of the Season or Team of the Season, often because they don’t realise that the award describes the guy that most opponents didn’t want to play against, or who actual professionals believed was the most valuable player - not who was objectively the ‘best.’

Fan votes are notoriously swung either by mischief (hello Lewis Mansell), by caprice, or by simple democratic expedient of ‘the fans who are most numerous’ (or are best mobilised online)1.

Sports Personality of the Year is not really about ‘the best personality’ and more about who captured media imagination.

And so I fully expect this series will annoy folks. I’m taking a particular lens (three different lenses actually) that sort of illustrate my point. How you look at the question influences who wins. Pick a particular metric, and it’ll show you who the Team of the Season are. Pick a different one: get a different team.

In the previous post of the series - before we signed two players (I’ll write about them at some point soon) - I picked out our three left wingers for the Teams of the Season.

As with the defence, I think it’s worth splitting out left and right sides. Thistle were unusual this year in having a strong, traditional winger on either flank. That was part of our strength; teams couldn’t focus on just one wing. But even then, Fitzy and Chalmers are not the same player - they offer different patterns and different styles.

So who, then, will fit into our three Teams of the Season on the right flank?

Impact

Again, a cluster of wingers all put themselves in contention for the impact team. Of the 11 players who played 1000+ minutes and spent time on the right wing, five scored 14 or better on the impact score. Two scored 16 or better.

Remember, this is a measure of players whose presence coincided with his team’s outcomes being measurably better. Improved xG and reduced xG against. Better goals return. Fewer goals conceded. Improved individual output in this role.

It’s not to say that this player is necessarily the best in and of himself (though often he is), but his production when he’s on the park correlates with better team-level outcomes than when he’s not on the pitch - and better than it would be with a ‘normal’ player in this role swapped in.

Interestingly, this time, the best impact was a less vaunted player - Ethan Walker of Ayr United. Walker is a slightly unusual player to have in the ‘winger’ set. His superpower is less his carrying or crossing (though they’re both fine), and more his creative, penetrative passing. He does really well putting the ball within 20m of the opponent's goal, and he gets into the box well himself, receiving a lot of passes and finding a shot without needing lots touches in the box. He opened the scoring several times this season.

In a lot of ways, he’s more of a traditional attacking midfielder, who drifts across the frontline behind the striker. He doesn’t cross or dribble as much as most wingers, but he’s great at generating space and passing where it matters.

However, he’s really more of a second striker than a true wide player - though I quite like the asymmetry of having someone play out wide who’s naturally a 10 (especially if you have a good right back overlapping), I think it’s maybe too much of a stretch to have him as our RW here. So though Walker was very high impact, the true impact right winger is Logan Chalmers.

Chalmers struggled to recapture his early season vibrancy in the last few months of the year - especially after a difficult injury. But over the season, his data profiles as the second best attacker in the league, behind McPake.

Whereas only 20% of Walker’s events happened on the right flank, over 50% of Chalmers’ action happened there.

Chalmers is what I’ve taken to calling a ‘kraken’ this season - a version of ScoutedFtbl’s ‘luxury’ player archetype (which is not to say that that he’s something you can’t really afford - just that someone whose main exertion and impact happens when you’re in possession). He’s a player you ‘unleash’ (“Release the kraken!” geddit?) when you’re on the ball. He’s in the 97th percentile for attacking actions (shots on target, dribbles completed, accurate crosses) - meaning only 3% of players get more - but only 19th for defensive actions (81% of players contribute more).

He’s in the top 5% for events in the attacking third, and for generating expected assists. He scored 3 opening goals, 3 ‘winning’ goals, and 2 other goals that changed games from a losing game-state to drawing, or drawing to winning.

2-3 times a game, he would carry the ball significant distances towards the opposition goal.

His shooting from mid-range, especially early in the season, was deadly.

That line of red dots along the edge of the box is telling. He provided a threat from that distance to counterbalance the close-in threat of the likes of Watt and Samuel.

Chalmers’ season was strong all round as a wide player - shooting, dribbling, crossing. Overall, he’s the impact right winger.

Undervalued

Moving on then to the undervalued team - the player who has a good impact that is disproportionately high given his value.

For this, we’re going to pick Matty Todd. Todd was one of the sleeper players of the season, really. He didn’t start otu in the main slot at Dunfermline, and he had to moonlight in various positions - right wingback, left wing, and so on. But over the course of the season, he ended up being one of the most potent creators in the league. Loads of dangerous, creative passes, good crossing, and a lot of end-product assists.

Todd is likely to be in the Premier League after eight seasons at the Pars, choosing to leave them now his contract’s expired. It’ll be interesting to see how he does against tougher fullbacks, and what his settled position is (it seems likely he’ll land at St. Johnstone). Over the season, his impact was definitely undervalued.

Wildcard

I described in my left wing post the idea of escape velocity - our winger ‘wildcard’ metric. It measures a player’s ability to attract attention like a gravitational pull, and then get the ball successfully out of it and into space by passing or carrying.

Chalmers scores high for this because he was strong both as a carrier and a passer.

But the second place - and our wildcard pick - is Taylor Steven, who yesterday signed a contract extension at St. Johnstone. I’m a big fan of Steven’s (and Simo obviously is too). He’s not always a true right winger, but he’s very capable out on the right flank. He had a great first half of the season at Arbroath, and then a solid return to Perth after being recalled.

He gets into an unexpectedly high number of offensive duels - much like McPake, though at a slightly lower volume - and wins them well. His physicality means he can withstand the gravitational forces of opposition attention. He gets fouled a lot which is a good sign that he knows how to draw defenders in - even if he can’t beat his man, he still finds a way to retain possession. He provides a lot of second assists - the pass before the assist - and a good volume of smart, line-breaking passes across the goal, as well as through balls. His technical passing ability is very strong, and whilst he’s not been St. Johnstone’s main outlet, he’s a great long-term option. Steven, then, is our wildcard right winger.

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left), Ben McPherson (right)

• Anchor midfielder: Charlie Gilmour

• Wingers: Josh McPake (left), Logan Chalmers (right)

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left), Craig Watson (right)

• Anchor midfielder: Oisin Smyth

• Wingers: Michael Garrity (left), Matty Todd (right)

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left), Shea Kearney (right)

• Anchor midfielder: Jason Holt

• Wingers: Aidan Fitzpatrick (left), Taylor Steven (right)

The players who were in scope for the 8s (a relatively big pool of 75ish) had a lot of crossover with other positions. Some of them were players who were also in scope for the anchor midfielder (think McAlear and Stanway), but are more impactful further forward. Some were central forwards who can drop further back (Euan Henderson, Andy Tod, etc. etc.). And some - who I’ll cover in a moment - were wingers.

Wingers and wide attackers are interesting in this league. Some clubs play with a more narrow attacking three midfielders rather than outright wingers (Marshall and Dow - and earlier Steven - at Arbroath for instance). Some of them are technically wingers, but they drive inwards (‘inverting forwards’) like Tod and Morrison at Dunfermline, or McPake at St. Johnstone. So the crossover between the attacking unit is much more fluid this season. It’s harder to badge players as true wingers, especially when clubs run a leaner squad. Wingers can also appear in attacking mid positions, or up front, even within the same game.

All this to say - I decided to pick the wingers for the team first, rather than the 8s, as the pool of players is much leaner, and the true wingers are more likely to be ‘belong’ in that role, and only moonlight in the other places. That way, when we circle back to the more advanced central midfielders, we won’t be picking players who are better positioned out wide, pillaging our rarefied pool.

As with the full-backs, I’m going to split this over two posts - one for the left sided wingers, and one for the right, because - as I said in the post on the full-backs, asymmetry is more of a feature in the Championship these days, and left and right sided players are not necessarily mirrors of one another (despite teams like Thistle causing havoc by swapping players from one side to another every so often).

So, then, to perhaps the easiest pick in any of the articles…the left winger. I’m selecting from a pool of only 10 players who are ‘true’ wingers - mostly LW rather than wide forwards or wing backs - who have played at least 1000 minutes in the league.

Impact

There’s not really much debating this one.

It’s Josh McPake.

McPake has had a standout season, possibly one of the best in the Championship’s history. He managed over 100 shots, opening the scoring several times, and scoring the winner on six different occasions. Eight times, his goals turned losing positions into drawing game states, or draws into winning positions (not including the opening goals), flipping games on their head again and again. He averaged more than 20 offensive duels a game (roughly twice as many as any other winger), winning more than half of them. He overperformed his xG by around 5 goals, and his shots typically improved the xG of a chance by 9%. He got back and contributed 4 or 5 defensive duels per game, and carried the ball significant distances 5 times a game. He attempted more than 11 dribbles per game, and one more than 60% of them and crossed into the heart of the six-yard box more than 1.4 times per game. He drew 2.6 fouls per game and converted more than 15% of his shots on target.

These are all just numbers - you could list them off ad infinitum.

But it’s pretty hard to argue that the most impactful player in the league, never mind as a left winger specifically. He’s not a traditional ‘chalk on boots’ winger - more of a modern hybrid who cuts inside, drives for goals, makes the opposition play on the turn, and disrupts patterns in the defence. He carries, he creates space for others just by his presence, and he shoots well. McPake is the impact player.

Undervalued

Let’s then look at who was the most undervalued left winger in the league.

There were several players who all scored high for this - Euan Henderson at Airdrie (another crossover left winger who can also play in a more central attacking role, or upfront, with really strong underlying numbers outside of his raw impact), Murphy at Ayr United, and Matty Todd at Dunfermline.

But the winner here is one of my favourite ‘data view’ players - Michael Garrity at Morton. Garrity gets behind the opposition defensive line with almost 50% of his touches, which is an extraordinarily high number, helping push defences backwards. He wins almost 60% of his attempted dribbles, and makes a lot of crosses - at least one or two of which per game land in the most dangerous areas.

He generates more successful attacking actions (accurate crosses, dribbles won, shots on target) of any player on the list other than McPake.

Garrity is young - just coming into his peak years - so he has plenty of time to grow. He’s the undervalued left winger.

Wildcard

The final team is built up using wildcard metrics. For the wingers, I’m going to use a new measure I’m working on. For now, I’ll call it ‘escape velocity.’ The best wingers tend to attract attention from their opponents - close marking, physical duels, attempts to unsettle the baller. To excel, you need to be able to attract that attention like a gravitational well, and then figure out a way to escape. To get out of the pressure, to move the ball onwards to the goal, despite the collapsing cloud of opponents around you.

Hence: escape velocity.

This is a measure that takes into account a player’s gravitational attraction of attention, but then also his ability to get the ball out of that either by carrying it himself, or by passing it smartly to a team-mate (who is now in space due to the opponents drawn in by the high gravity winger’s pull).

The escape velocity relies on evidence of either exceptional ball carrying, or elite passing under pressure, or both.

Josh McPake comes out top of this list again (natch), with the best carrying in the league under pressure, and the second best passing. As he’s already made it into the impact team, let’s skip to number two…

…who is Aidan Fitzpatrick.

Fitzy has had a stellar season. Whilst he’s obviously a great ball carrier (he’s second only to McPake in this), his superpower this season has actually been his passing under pressure. Driving into traffic, then cutting the ball across the goal to a team-mate. Creating space passively for others by drawing opponents in towards him and then laying the ball off to Smyth or Stanway or Watt or Reading or…

Fitzy’s passing out of traffic is better than any of the other left wingers in the list, and his carrying in traffic has been pretty stellar this year too. No-one else on the list generates as much xA per 100 passes as Fitzy (though Dow comes quite close). If he decides to leave Thistle this summer, it’ll be a big loss - Wilson has clearly understood how to get the best out of him, and surrounded him with the right players to help him sing. It’s hard to capture quite what we’ll lose - one of our own for sure - but if nothing else, he’s clearly one of the best wingers in the league.

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left), Ben McPherson (right)

• Anchor midfielder: Charlie Gilmour

• Wingers: Josh McPake (left)

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left), Craig Watson (right)

• Anchor midfielder: Oisin Smyth

• Wingers: Michael Garrity (left)

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left), Shea Kearney (right)

• Anchor midfielder: Jason Holt

• Wingers: Aidan Fitzpatrick (left)

It was interesting to me how much the team’s style changed when one player was swapped in or out. When Ben Stanway played at the six, we often had more dynamic attack, but it was more chaotic - I called him the ‘chaotic catalyst’ in that particular role (this was specifically to do with when he was playing in the anchor ‘6’ position). Oisin Smyth was more of a ‘cultured conductor’ who dominated possession. Luke McBeth, on the other hand, was the ‘intense disruptor’ who kept the out-of-possession tempo high and aimed to pass directly up the pitch.

There aren’t many other roles on the pitch that change a team as much as this anchor midfield slot.

So as we continue building our three teams of the season using the skewed lens of a single metric for each team, the feel of the three sides - the balance of attack and defence, the preference for high or low possession - will change depending on who comes out on top.

Impact

The first of our three teams is based on impact - players whose presence improves the output of their team, contributing to a better goal difference (and expected goal difference) when they are on the pitch than when they are off. It’s not quite the same as saying the ‘best’ player - just the one in this role whose presence is most positively influential to his team.

This position is one of those that can genuinely impact both ends of a side’s output - disrupting opponents and reducing their capacity to create chances on one side of the ball, and helping in buildup, contributing to chances on the other side.

There are four players who played in anchor midfield positions this season (with at least 1000 minutes) who were clustered together in the top right of the chart above - all of them had significant positive impact on their teams. Two of them could be argued to get the ‘red’ dot of the most impactful player. McAlear at St. Johnstone is one of the most compelling players in any position on the pitch, putting up big minutes and with impact to spare.

But the player who edges McAlear (and the other two) out is Charlie Gilmour at Dunfermline. Gilmour isn’t necessarily the most eye-catching player in the Pars team, but he’s been exceptionally important to them this year. He passes well, he gives strong aerial protection in front of the defence, and he is great at drawing fouls from opponents, relieving pressure on his team at the edge of their own third. His passing accuracy is over 85%, and over 75% passing into the final third. He breaks up play well and recycles the ball incredibly reliably - but he’s not cautious in doing that. In lots of ways, he’s the heartbeat of that team, and has had a huge impact on the pitch.

Undervalued

The cluster of four players at the top right of the chart above all get around 14 or better in terms of impact. But when you then control for their market value, two of them are north of 18. These two are players who are both impactful, but whose impact is disproportionately high considering the value that Wyscout give them.

One is another player who spent this year at Dunfermline - Chris Hamilton (who has recently driven north to Dingwall).

The other - the highest for impact by value - is Oisin Smyth. Smyth’s data impact on Thistle is high, but it was much higher than you’d expect it to be given his market value. He outperformed all expectations, controlling the midfield in game after game. His ability to release runners with passes from deep to the wide channels was a sight to behold. He managed almost 60 passes a game, with an accuracy over 80%, and contributed an unusually high number of successful attacking actions for a deeper midfielder. His average pass was over 20m, with a successful long pass rate of over 50%, and 75% for progressive passes that moved the ball significantly towards the opposition goal.

Smyth - although he was widely recognised as a great player - was also the most undervalued contributor in this anchor midfield position. He was Thistle’s cultured conductor at his best.

Wildcard

The third of our teams is based on a wildcard metric. For each role in the team, I’ll pick a role-specific metric which allows us to find a player who stands out from the rest. Anchor midfielders come in different flavours - distributors (who co-ordinate play by picking out team-mates), screeners (whose main role is to protect the line behind them), quarterbacks (who control the start of the buildup by making passes that launch attacks), safety valves (who release the pressure their team faces by finding the right outball and breaking up play).

For this particular wildcard, I’m going to look for players who are important to the buildup phase of their team.

I’ve created a metric that combines a player’s contribution to the earlier stages of buildup1 - because they make passes that often lead to assists - with his contribution to bring attacks to completion as well. A number above zero means the player is above average over these two disciplines - either because they are exceptional at one, or they are (at least) pretty good at both.

Top of the tree here is Ryan Duncan, who was on loan at Ross County from Aberdeen this year. Duncan was well above average at both early and late buildup contribution. He’s marked in Wyscout as playing some games in the centre of midfield, and so I’ve left him on the chart for completeness. However, he’s actually more of a wide midfielder rather than a deep anchor (you can see his heatmap below), and so I don’t think he’s the right call for the team here.

So the winner in this, then, is Jason Holt. Holt contributed a lot in early buildup - he’s especially good at getting ‘third assists’ - the pass before the pass before the assist (one for the woke stats mob there). And he’s also pretty good at getting passes into the final phase too.

In a lot of ways, Holt was probably one of the - if not unsung, then at least only moderately serenaded - heroes of St. Johnstone’s season. He and McAlear were essential at controlling games, allowing the likes of McPake and Gullan to excel in front of them. His sideways passing was 96% accurate, and passes into the penalty area 68% accurate. He held the ball well under pressure, drawing a good number of fouls per game. His passing was effective at creating shooting opportunities. He may not have commanded the headlines (at least for opposition fans), but he certainly contributed massively to his team’s success.

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left), Ben McPherson (right)

• Anchor midfielder: Charlie Gilmour

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left), Craig Watson (right)

• Anchor midfielder: Oisin Smyth

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left), Shea Kearney (right)

• Anchor midfielder: Jason Holt

Start writing something, thinking it’s going to be relatively quick and to the point, and then realising there’s more to say - more nuance, more context, more edge cases - far more to say than I realised. The process of research helps to shape the story, but so does the process of writing itself. Which is, of course, why you really ought to leave yourself time to redraft at least once. Ah well, it’s good to have a dream.

So I decided to bail out early in this series. Rather than attempting to overload any one post, I figured it was better to allow each set of players to breathe a bit.

Fullbacks is a case in point. Conventionally, you’d expect the only difference to be whether the players play on the left or the right side of the pitch. But in practice…anyone who’s watched professional football evolve over the last year or two (I just can’t bring myself to use the phrase ‘tactical meta’ with a straight face. Even I have limits.) has seen that often teams use their left and right back differently.

Thistle, for instance, use Paddy Reading and Ben McPherson / Ethan “The Major” Ingram / Cammy Logan differently. Of course, this is somewhat influenced by the available players, and broader team dynamics. Paddy Reading’s superpower is, in many ways, less about his individuality and more about how he draws his team-mates out. I’m of the opinion that O’Reilly, Smyth, Stanway, and Fitzpatrick all had better seasons because of Reading’s input. He knew when to cover, when to overlap, when to step inside, when to push. He tucked into midfield to allow Smyth and Stanway to bomb up. He edged forward to give Fitzy more support when he ran for the corners. He stepped over to help out Ashcroft and O’Reilly’s high line. He was rarely the main outlet of the attack, but he gave at least four other players the opportunity to contribute individually.

Ben McPherson, on the other hand, offered an attacking threat himself - sometimes playing more like a right wingback.

Reading spent a lot of time in the middle third of the pitch, and more in the defensive third than McPherson did. McPherson is much more likely to curl round into the right edge of the box towards the end of an attack in a way that Reading won’t.

Teams often use this asymmetry these days. That’s my point.

And so it’s worth considering the right backs in the league on their own, in isolation from the left backs.

Impact

The first of the three teams we’re building is based on impact. Finding players who make their team better. Take a high-impact player out, and his team is less impressive without him.

For this position, we have 24 guys who have technically played at RB or RWB this season, and who have broken 1000 minutes in the league. The eagle eyed among you will have noticed that there are two players further right than the one I’ve highlighted. This is not just Thistle bias (I know, I know - you believe me, but thousands wouldn’t). I’ve taken a call to leave out the top two for impact because I don’t think either of them are really fullbacks (though they can play in that role quite effectively) and I’m aiming to build a team that plays in the Thistle 4-3-3- shape if at all possible. Chilikoa-Mullen and Akil Wright are both excellent defenders, in different ways. But I think if you had either in your side, you’d choose to use them as a CB (maybe in a back three) rather than as a FB.

Which means the highest impact RB is Ben McPherson. It’s been a great season for the Thistle player. He’s had a challenging end to the year, trying to recover from an injury and not quite capturing his pre-knock form in the playoffs, and he’s had very able deputies like Ingram (who enabled a memorable couple of games with McPherson as a goalscoring winger) and Cammy Logan who started and ended the season well.

McPherson is a good crosser, and his crosses tend to go deep into the box. He’s a great ball progressor and a very strong technical passer, as well as completing almost 65% of his dribbles man-on-man. He is also exceptionally strong at finding space to receive passes from team-mates. McPherson is often the guy who receives the out-ball when more central colleagues are under pressure, which allows others to move forward.

McPherson has been the most impactful true right-back in the league this year.

Undervalued

Who, then, is most undervalued? Remember, this combines both impact, and how strong the impact is relative to the player’s perceived financial value. We’re seeing a pattern emerge here that generally this tends to go to a player who is either close to running down his contract, or who is at the more experienced end of the age range - players who wouldn’t necessarily give you big resale value, but from whom you could get a season or two of high output relative to the outlay for them.

In this regard, the right back of choice is our first Arbroath player - Craig Watson. Watson is a more defensively-minded flavour of RB. He doesn’t bang in the crosses the way someone like Kearney or McPherson or Doherty might. But he’s a great interceptor - cutting out passes before they turn into something dangerous. And he’s a good threat passing into the box, making more non-cross passes into the danger zone than most other RBs. He’s got a not-bad ability to change pace when he’s carrying the ball. He’s got one of the better rates of defensive actions in the league.

Watson’s not going to contribute 20 assists a season, but he’s going to keep your right flank strong, making it very hard for the Fitzys of the world to bomb forward unimpeded. Which is a very underrated contribution. Hamilton at Dunfermline (recently moved to Ross County) gets a similarly high score here. But Watson is the winner.

Wildcard

As with the left-backs, I’m looking here for players who consistently move the ball significantly up the pitch, either by passing and running (or both). On the whole, the left side of the defence seems to excel at this more than the right. Of the five players who get a score of 1.0 or greater (more than 1 standard deviation above the league mean), none of them play on the right. In fact, only four of the eleven who score above 0.5 are right-sided, and two of them (JCM and Doherty at Raith Rovers) are only sort-of-fullbacks (and Forrester is a bit of both sides).

There are probably a lot of reasons for that (see “prevailing asymmetrical tactical meta” and other suitably self-important tactico terms). But there are two players who are well above average for this on the right - Shea Kearney at Dunfermline and Adam Forrester at St. Johnstone. Because Forrester was a bit of a late entrant into the side, and swapped between both flanks, I’ve pushed in Shea Kearney for this team.

Kearney isn’t the most compelling all-rounder…yet. But I think he’ll come good in time, and he’s already trusted by Lennon on the opposite flank to Fraser. In my internal grading system that I use for a quick-look at players, he doesn’t score below a C in any particular metric.

He looks really good going forward (high cross accuracy, lots of dangerous balls deep into the box, wins around 70% of his dribbles attempted), and has a big upside, though he’s a bit more suspect in duels and defensively, especially in the air - Thistle have had some joy getting in behind him. But playing in a unit alongside Ashcroft … probably not too big a risk. One small note though - whilst he has a high volume of ball progression, his accuracy over longer distance isn’t as great as it could be. As I say - not the finished product, but a player with a good upside in his second season. Which is, probably, what you want from a wildcard team.

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left), Ben McPherson (right)

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left), Craig Watson (right)

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left), Shea Kearney (right)

So far, we’ve added three players to each of three teams: ‘keepers and centre backs for a team based on impact, on value for money, and on a role-specific set of wildcard metrics.

Next up: a pair of posts on full-backs. And first up - the left side. I’ve limited it to the 18 players who played at least some minutes at LB or LWB and had at least 1000 minutes played over the course of the season.

Impact

The first of the three teams is built on ‘impact'.’ Importantly, this is a blend of the general production stats for players in this role, alongside a replacement score - how much better was the player’s team when they were on the pitch compared to the times they were off, and how much better is that compared to a relatively average league player in this role.

This is important to bear in mind because it’s a qualitative measure of how much better a player makes his team - or at least how correlated a player’s presence is to the team being better. Mostly, it’s a good way of spotting the high-performing players (so far, it’s picked out Toby Steward, alongside a defensive line of O’Reilly and Diabaté which seems fairly accurate). That’s not always the same thing as ‘the best player’ as far as fans are concerned. But it does show whose presence coincided with a team’s best performances.

By this metric, the most impressive left-sided defender was Morgan Boyes of St. Johnstone. I mentioned him earlier in the season as a bit of a unicorn in this area.

I know St. Johnstone fans are not that convinced with him in that position (he can also play in central defence). But he demonstrably is on the pitch when they had the best performances. I’ve long been a fan of Boyes for being one of the best in the league at building up from deep, and whilst he’s not a traditional left back, he’s still pretty impressive, and he’s been able to keep St. Johnstone levels up when he’s on the pitch.

Undervalued

Next, then, the left back for the ‘value for money’ team. Remember, this player is the one who gets the best balance of impact (by the same measure above) held against how much Wyscout believes the player would be worth on the market. This somewhat favours players with shorter contracts or who are a bit older, as they tend to have slightly lower market values. In short, the best players by this metric are generally undervalued.

As with other positions, this is quite a good way of getting players who are unexpectedly performing well. This time, the metric picks out Raith Rovers’ Lewis Stevenson (quite a long way ahead of Morton’s Ballantyne). Stevenson, in his late 30s, has turned in around 2.5k minutes this season, which is pretty impressive, and has returned impressive numbers for winning duels (over 60% of defensive duels, overy 50% of aerials), some decent numbers for changing pace whilst carrying the ball, and generally excellent passing numbers. All in all, he’s done remarkably well for a player who’s … not necessarily in the early days of his peak.

Wildcard

For our wildcard for FBs, I’ve decided to use ball progression - a player’s ability to move the ball significantly and successfully up the pitch towards the opposition’s goal. The ability to do this is very well correlated with team success, especially from deep. Some players do this by passing, some by carrying the ball. But doing it well and often is the engine room of a high-performing Championship team.

I’ve measured a combined score across the two different ways a player can progress the ball - above zero means the player is above average for actions that progress forward.

Morgan Boyes is the strongest player in a while at this particular discipline. But after him, there are two players who are neck and neck for the wildcard team - Matty Foulds and Robbie Fraser. I’m a massive fan of Fraser, and would love to have him at Thistle one day. But if we’re sticking to the data, Foulds just edges it.

It’s an interesting feature of St. Johnstone’s season that they couldn’t really settle on a left back. Foulds, Boyes, and Forrester all had time there, and the fans weren’t really convinced with any of them (which is, of course, why you should never rely on a single metric to make recruitment or selection decisions!).

Foulds, though, progressed the ball almost as well as Boyes - at least in terms of passing (Boyes also is well above average for running the ball, which Foulds doesn’t do quite so much). He managed 2.92 attacking contributions (successful dribbles, completed crosses, or shots on target) per game, and passed into the penalty area more than five times a game on average. That’s pretty impressive. And he just edges Fraser overall as a progressor. So the player for the Wildcard Team is Matty Foulds.

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

• Fullbacks: Morgan Boyes (left)

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

• Fullbacks: Lewis Stevenson (left)

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

• Fullbacks: Matty Foulds (left)

Still, that’s what we do around here. So let’s add to our nascent three teams: one built on impact, one on value for money, and one on a role-specific wildcard metric.

Last time, we found our ‘keepers. Now, let’s protect them with a pair of central defenders.

Impact

As I explained last time around, this is a measure of a player’s relative impact on their side’s performance. A high-impact player sees his team get a much better number for (xG - xG against) when he’s on the pitch compared to when he’s not on the pitch. And then that number is compared to the same number for a fictional ‘average’ player in the league in the same position.

Effectively, if you swapped each player out for a normal Championship player, would the team get a worse, a bit better, or a lot better?

In actual fact, my metric takes into account other things as well (though it’s anchored against this ‘Wins above replacement’ metric I’ve described). It accounts for a player’s output relative to what you want from someone in this role. But overall, each player is given 0 to 20 to grade how much of a positive impact they’ve had. 10+ is good. 15+ is excellent.

The two high-impact defenders, highlighted in red in the chart above (who have played at least 1000 minutes) are Dan O’Reilly and Cheick Diabaté. This, at least, passes the eye test - both of these guys could be argued to be in the top five defenders in the league.

It’s also interesting to me that O’Reilly scores so high for impact. The flip side of having a high score here is that if you take this player out, the team is more likely to struggle. As good as Cale Loughrey has been (and I think he’s been excellent given the step up he’s needed to make), losing O’Reilly was hugely impactful (in a negative sense) in the last two games. It made it harder to have dependable buildup patterns, and it made it more likely we’d struggle in holding a high line.

Diabaté has been excellent too - a highly reliable defender, with aerial presence, and good passing instincts.

So it feels sensible that these two are our impact defenders.

Undervalued

Onto our second measure. Defenders who are undervalued. In other words, those with a good impact score that is higher up the league than you’d expect, given what Wyscout thinks they’re worth on the transfer market1.

One of my arbitrary rules is that a player can’t be in more than one of my teams. So, although Dan O’Reilly is top for this metric too - he overperforms his impact given what his value says he’s worth - he misses out of this team (I’ve given preference for Impact).

The two players who make it into this team (in yellow above) are a bit more left-field: Callum Fordyce of Raith Rovers, and Jackson Longridge of Morton.

Fordyce is one of those players who fans of his own team tend to rate quite highly, but opposition fans don’t know that much about him. He’s a very accomplished passer of the ball, and a great presence in Raith Rovers’ defence. He only managed about 50% of the minutes this year, but he’s still an impressive player. I mentioned earlier in the year that he was close to being in an elite group of players who manages a high rate of interceptions, wins a lot of his duels, and contributes a lot of progressive passes (he’s slightly lower on the interceptions this year). He’s also, apparently, Rhys McCabe’s cousin2. So there you go.

Longridge is an interesting one. He’s an OK passer (though not as good as the others), but his defensive output is significant. He wins over 60% of his defensive duels, and contributes a lot of defensive actions each game.

These two, then, provide a different sort of defence in front of Harry Stone. Not necessarily the culture and presence of O’Reilly and Diabaté. But a compelling defence nonetheless. This is what the undervalued metric should uncover.

Wildcard

Finally, then, a wildcard role-specific metric. For this one, I’ve generated a single number that represents the proactive defending instincts of each centre-back in the list. It combines how good each player is at contributing sliding tackles and interceptions - proactively intervening in the opposition attacks, without waiting for the other side to come to them.

It’s quite abstracted away from a single thing you could measure in the real world - but it helps to see who the most proactive disruptive defenders are.

Two defenders are miles ahead in terms of tackles alone (Akil Wright and Iain Wilson - though of course Shug isn’t really a CD in the true sense). But when you wrap it together with interceptions, and account for how much possession a player’s team has (so that you don’t just pick out the teams who have a lot of chances to defend because they’re overall worse), you get a more nuanced picture.

The two players to come out here are John Tod and Lee Ashcroft. Tod is one of those players who I think is underrated - Chilikoa-Mullen and Ngwenya tend to take more of the plaudits. But Tod’s numbers are pretty impressive for such a young player, alongside much more experienced defenders. And, of course, Ashcroft has been a huge part of our success this year. Although other players will be missed when they leave Thistle, our two central defenders are, I would argue, the reason we’ve had such a good season.

Ashcroft’s superpower is intercepting the ball - he’s exceptional at it (not just this season) - I mentioned him last Summer in the same vein (and Longridge too, as well as Wilson and Charlie Telfer from further forward).

The State of Play

Here, then, are the teams so far:

Team 1 - Impact

• Goalkeeper: Toby Steward

• Central Defenders: Cheick Diabaté and Dan O’Reilly

Team 2 - Undervalued

• Goalkeeper: Harry Stone

• Central Defenders: Callum Fordyce and Jackson Longridge

Team 3 - Wildcards

• Goalkeeper: Calum Ferrie

• Central Defenders: John Tod and Lee Ashcroft

Next up (probably): wide defenders!

Overall, it’s been a fabulous effort by the team. Of course, I would have loved it had we managed to get past the final hurdle. But given that this was expected to be a rebuilding season, when we reset the team, brought the age down in the squad, and set up for years to come, finishing second and getting to within one half of the playoff finals of promotion is pretty impressive.

It’s been a while since I’ve enjoyed a season as much as this one. The culture of the team, the exciting and varied style of play, the generally likeable nature of the players, the atmosphere created by the fans - it’s been great. Not perfect. But pretty close nonetheless.

Eventually, I’m going to get around to looking at the squad and what we’ll need to do for next season to go again. But for now, I’m planning a series of relatively short posts reviewing the championship season.

I’m going to build three teams of the season, position by position. The teams will be in the Wilsonball 1.0 shape - 4 at the back, one defensive mid, two running 8s, two wingers, and a central forward through the middle.

The first team is going to be based on "overall impact”. Basically, I’m going to use an updated version of my WAR score that considers how much better a team is when a particular player is on the pitch, to when he’s not, and then compares that impact to how the team would do with a reasonable replacement (somewhere in the middle of the Championship). Players who score high on this have a disproportionately positive impact on their team’s expected goal difference.

The second is a team of ‘undervalued’ players. This one compares the player’s impact, and his market value based on Wyscout’s estimation. A player who has a good impact and a low value will be ‘undervalued’. A player who’s worth a lot but doesn’t improve his team much will be overvalued.

The third team will be based on a wildcard metric that I’ll decide on role-by-role.

The only rule will be that the chosen players for the three teams can’t be the same player. So we’ll end up with three distinct teams of the season, from slightly different angles.

Let’s start, then, at the beginning (for it’s a very good place to start). Who’ll be in goals?

Impact

The impact keeper, probably unsurprisingly, is Toby Steward. Steward returned an impressive set of clean sheets this year, and consistently impressed for a team that topped the table from the first day of the season. If you swapped Steward out for a ‘normal’ keeper in the league, St. Johnstone would have been noticeably weaker at the back.

I wrote about Steward when I was looking for a replacement for Myles Roberts last season - I’ve liked him for a while. He’s a good distributor, a strong frame, impressive shot-stopper, and has stepped up well to a higher league this season.

Second, close behind Steward, was Josh Clarke. But on this metric, Steward is the one to take the gloves.

Undervalued

Let’s spin it around a bit now. Instead of just considering the impact of the player, let’s add in his market value too. Steward still scores high, but this time, he’s overtaken by Airdrie’s loan keeper (and onetime Jag) Harry Stone.

Stone improved Airdrie quite significantly when he joined, after they’d cycled through a few young and loan keepers. He’s never quite locked down a team despite several loan spells, but I’d wager that this half-season should be enough to convince a Championship team to take him on again next year.

Some surprising players rise up this chart - mostly because of a low market value for a Championship keeper. Steward still does well here, and Billy Terrell at Dunfermline comes up impressively, too. But Harry Stone is the best balance of performance and value.

Wildcard

The final ‘wildcard’ team, then, is based on the main metric that matters for keepers: how many goals they keep out.

Calum Ferrie wins this by some distance. If you compare the xG he faced over the season with the actual goals he conceded, he kept out far more than you’d expect him to. We’ve all known for several seasons that Ferrie is a great shot stopper. This year hasn’t been his best overall at QP, but he’s still kept out more goals than anyone else in the league - more than twice Josh Clarke in second.

As an aside, it’s worth keeping an eye on this metric. Especially in certain leagues (like the National League in England), this is a very strong way of finding players who can perform at our level.

The state of play

So then, we have our first three players into our data-only team of the year (I know this isn’t the best or only way to make this sort of team, but it seems like it’s appropriate for my vibe!).

Based purely on impact: Toby Steward

Based on who’s undervalued: Harry Stone

Based on a wildcard metric: Calum Ferrie

Tune in next time for our defence!

Thistle have been extraordinarily resilient this season - going all the way to the final whistle. We have scored 8 goals in added time at the end of the game this season - more than any other side.

In fact, only Dunfermline have managed more than 2 across the rest of the league. And last night (and on Tuesday), we showed that resiliency again - coming from behind in both games, and going on to win the game in regular time last night. What a rush. And what a turnout in the ground, full of sound and colour and joy and shared experience.

I’ve begun to think ahead to how we figure out if a player would fit in at Thistle. One way or another, there will be some gaps we need to fill in the Summer - either as players move on, or as we need to build a bigger squad for a higher league.

One particularly useful tool is the idea of the ‘archetype’ that Scouted Ftbl use to describe players. You can read more about them here - it’s a way of understanding what a player excels at - how he influences the game, his area of influence, and so on - that goes beyond positions. They had an excellent podcast this week looking at the future of midfield that covered some of this too.

The idea is not to fit every player into a bucket, so much as to find excellent examples of particular types of players in the modern football pattern. Some teams play in a way that relies on specific types of player, and the ‘best’ example of this is the archetype. I’ve said before that I think this is helpful and connects with the sort of storytelling that humans are better at engaging with.

I’m beginning to explore what sort of ‘archetypes’ Mark Wilson wants to play in his teams, as that will help us identify players who would fit well - and excellent players who wouldn’t necessarily excel in his system, which is equally as important. Just being a ‘good’ player is not enough.

For example - we need defenders who are good at the sort of things Lee Ashcroft is good at for our early buildup (comfortable sitting high, good accurate long range passer, able to physically dominate his opposite number, reads the game in front of him extraordinarily well). And we need different sorts of attackers (less the Brian Graham high-xG lurker, and more either the sort of link-up forward that Watt provides or back to goal runner that Samuel typifies).

Spencer Mossman is great at coming up with pairs of attributes that show which players fit which archetypes. He had another great one this week, looking at wide attackers, which is a crucial part of this system.

I’ve plotted some players from several leagues onto the same pair of axes, trying to unlock which wide players are better at ‘picking locks’ and which are ‘roadrunners’. Intuitively, we all know that some wingers are dribblers, and some find open space to run into. But who matches which archetype most closely, and where do Thistle players sit?

What’s interesting to me on this chart is where Chalmers and Fitzy sit. They are more towards the roadrunner quadrant than the lockpicker - not seeking to beat a man over and over again, but instead to find space to receive long passes.

I’ve like Marc Emslie of Ross County a lot this season (on loan at Cove), and also Sole of Livi (at ICT). But in reality, they are less like Thistle’s wide attackers and closer to someone like McPake who wants to take on a man rather than find the space himself.

Much closer in type to Thistle’s attacker is another on loan winger - McLuckie of Hearts.

This tells us something about who would do well in Thistle’s system (assuming we keep playing a similar way next year). Wilson wants wide players who find space to receive the long passes that the likes of Smyth, Ashcroft, and O’Reilly ping around.

It’s also interesting that both of the players rise above the throng. There are more than 800 players on that chart, but hardly any get above 3 on the y-axis - and both Thistle wingers manage that.

I’m not going to write any more on this now - I’m off to enjoy a day reflecting on last night. But just to say that the type of player matters as much - or more than - his position. Positions can flow and change. The players we have next season - wherever we may be - need to be the right type.

Enjoy the weekend, Jags fans. See you on Thursday!

It’s playoff time.

As is often the case, a conversation earlier in the week got me thinking

On Draw, Lose, or Draw, we discussed the last four years of playoffs and what that means going into this year. After all, we remember the incredible vibes as Thistle almost drove to the Prem back in 22-23, with McMillan, Lawless, Tiff, BBG, et al. What was the characteristic of that year, and what was better or worse in other attempts? And are we in a better or worse place this season to make a run at getting promoted?

The first thing that’s worth saying is that, of course, this season is the first time in that stretch that we haven’t had to play in the first round - which has several benefits (the players are more rested, there’s more time to prepare and research, training is more controlled, and we get home advantage in the second leg, to name a few).

But what else can we factor into the start of this playoff campaign?

I’ve picked out a few areas that are worth paying attention to. In each case, I want to review a few things:

1. How has our ‘regular season’ pattern been and is it more similar to 22-23 (big vibes), or to 21-22 (too much too soon), 23-24 (worse than 22-23 but still pretty good), or last season (a step too far with a jaded, battered squad)?

2. What changed in each of the previous seasons between the regular season, and the resulting playoff run? This might give us a clue what we need to do to get over the line this time

Overall game balance

First, let’s look at our overall game. In each of the following charts, the yellow dot shows the ‘regular season’ and the red dot the playoff games1.

Area: expected goal difference. In other words, add up all the xG we generated and take away all the xG we conceded, and divide by the number of games played. xGD of +1.0 means that in an average game, we generated around 1 more xG in attack than we condeded to opponents. Over a season, this is a good indication whether the team should be winning games.

Regular season patterns

• In 22-23, we had an incredibly strong positive xGD even before the playoffs - between +0.7 and +0.8.

• 23-24 was still cautiously positive, 21-22 less so, and 24-25, we were basically break-even

• This season, we’re doing pretty well - the second most impressive xGD since we returned to the Champ. It’s not as high as the vibes season, so we’re starting from further back. The big issue has been the last couple of months, where we’ve allowed opponents to generate more chances against us.

Playoff trends

• In almost every previous year, we’ve gone up in the playoffs - generated a better xGD than in the regular season.

• In 21-22 and 23-24, it climbed almost to the level of the regular season in 22-23.

• Last year, with a more limp performance in the playoffs, though, we dropped off from a low start point to a negative xGD

So what?

• We’re starting from a good base, especially if we can improve our defensive numbers to where they were earlier in the season

• In order to get to our more successful levels, we need to lift this even more - more xG generated, less xG conceded. Ideally, we need that number to swing by at least +0.5

Area: Possession - this is pretty straightforward.

Regular season patterns

• We have the highest average possession of any season in the last five - only 22-23 was even close

• For the last two seasons before this, we have had under 50% of the ball - that has been decisively reversed this year

Playoff trends

• In pretty much every season, we end up with less possession in the playoffs. This is sort of expected - you’re playing against the stronger teams in your own league, plus a well-financed team from the league above if you get to the final, so it’s harder to dominate than it is agains the weaker teams in the Champ

• Normally, the drop-off is between 3 and 4 percentage points.

• In 22-23, this kept us just below 50% - this year, it would keep us just above

So what?

• The big thing that’s different this year is that we are set up to dominate the ball - to impose our gameplan on opponents. We can expect not to be quite as dominant, but that changes the feel of the matches

• More time on the ball is better for stamina - you don’t have to press intensely, you don’t have to focus on slimmer pickings going forwards.

• It’s encouraging though that we go into the playoffs with a much more similar profile to 22-23, and we’re likely to have a smaller drop-off because we only have to play semi-finals and (hopefully) finals, and not the QF.

Attacking

Area: Shot volume - basically how many shots on target we generated per game

Regular season patterns

• in 22-23 and 23-24 (the two seasons we were more competitive), we started with aa higher baseline - over 4 shots on target per game

• In 22-23, that was almost 5.5 per game

• This is related to the xGD, but is also a way of saying not just that we get chances, but that we actually force the opponents to do something about it - to block or save shots

• Last season, the rate of shots on target was much, much lower - only 3.5 per game. That is a measure of the fact that our attack had become collapsed down to ‘can BBG get a shot?’ by that stage.

• This year, we’re starting from a baseline between 23-24 and 22-23 - north of 4.5 per game

Playoff trends

• This as an interesting difference to what we’ve seen so far. In the ‘good’ runs, we lifted our attack noticeably - at least 1 or 2 more shots on target per game

• Interestingly, 22-23 wasn’t quite as much as a lift - in 23-24, we lifted the shots on target volume to higher than the previous year (the Blair Alston effect, obviously)

• In the less compelling seasons (21-22 and 24-25), the shots on target rate dropped off noticeably

So what?

• We’re starting from a good place - with an attack that generates a lot of chances, and forwards who turn those chances into shots on target

• If we want to make a run, we’ll have to lift that further, probably adding at least one or two more shots on target per game than our regular season average.

Area: Deep completed passes per 90. These are passes that put the ball to within 20m of the opposition goal. They are a dangerous sort of pass that creates chances for team-mates in the higher-percentage zone. Normally, a dangerous attack excels in at least two of three creative styles - crossing (think Harry Milne), running (think Scott Tiffoney) with the ball, or passing into dangerous areas for players to run onto. Thistle have been at their best when they generate a lot of these deep completed passes - they might be through balls, cut-backs across the face of goal, or other passes, so long as they end up within 20m.

Regular season patterns

• 22-23 is the only season we’ve averagedmore than 7.5 deep completed passes per game - this is the sort of pass that Stevie Lawless excelled at

• In every other previous season, we were 7ish or lower, and as low as 6 last year (this was one of those areas where we had to recruit in players and train tactics that maximised because we fell away massively last season, and it damaged our attack)

• This season, though, we’re starting from our highest baseline in five years - almost 9 deep completed passes per 90

Playoff trends

• It’s generally harder to make this sort of pass against more structured or tenacious defences - for instance, a well rested unit who’ve skipped the first round, or an intense presser like Livi last year, or a very deep defensive line like Dunfermline have reverted to recently

• In three of the previous seasons, the playoff rate stayed quite close to the regular season

• But in 22-23 - the big season - we catapaulted this number up hugely, from around 7.75 to 10 per game. That’s huge, and it was the reason our attack was so fun to watch. Again and again that year, we were putting balls deep into danger

So what?

• We’re starting from the best structural attack in five years. We’re not at vibes playoff levels, but we’re not far off it. It wouldn’t take much to rachet it up

• We probably need someone like Turner, Taylor, or one of the attacking 8s (like Stanway) to clock in a few more of these. But we’re structurally set up to do it in a way that we haven’t been since 22-23

• This is the single most correlated factor that we could easily replicate this year

Area: Getting the ball into the penalty area. There are two other ways to get the ball into ‘having a chance at a shot’ territory - running it in, or crossing it in.

Regular season patterns

• The more successful seasons had around 4-5 runs into the penalty box per game in the regular season

• We’re at about that level this year (a touch lower than 22-23, but still in that zone)

• Crossing is more volatile, but we are currently at the highest volume of crosses to the penalty area per 90 of the last five seasons

Playoff trends

• In 22-23, the big difference to the other years is that we ran the ball into the box more in the playoffs than we did the rest of the season (think Tiff, McMillan, et al bombing into the box)

• In all the other seasons, we’ve dropped off our running attack - sometimes by as much as 3 runs a game

• Cross volume also tends to drop a bit - in 23-24, we crossed in more in the playoffs than the rest of the season, but in 22-23, we effectively traded crosses for more runs. With the team we had that year, that made sense, as we had multiple players who could wreak havoc this way.

So what?

• We have the best crossing team we’ve had in years, and our running box entries are around as good as peak playoff seasons

• The way we succeeded in 22-23 was to cross a bit less and run a lot more in the playoffs - probably because it’s a way to unsettle teams that are tiring, and keep them unsettled. Chaos is good!

• That’s not to say that we need to do that same thing this year, but this year, we don’t have BBG to lurk in the penalty box to get on the end of crosses, so getting the likes of Stanway, Watt, Fitzy, Chalmers, McPherson, and co running at teams seems like a good way to replicate this form

Out of possession

The other side of the ball matters too - especially as you end up facing better equipped Prem teams, or teams with potent running and shooting threats.

Across several metrics, there are fewer clear patterns out of possession, though.

We tend not to change as much between the regular season an playoffs, and there are fewer patterns in terms of the regular season itself.

• We are winning a higher fraction of contested balls than in any previous season this year

• We are making far fewer sliding tackles (no Banzo, no Milne)

• We are recovering the ball more often in the opponent’s third, close to their goal much more

• We are pressing more intensely than any season since 21-22

• In most previous seasons, the out of possession numbers stayed quite consistent, other than 23-24, when we shifted to being more aggressive, winning more duels, recovering the ball more, making more tackles, and so on

Basically: at first glance, the most important thing to lift is our attacking output rather than our defensive numbers. It’s crucial to maintain our best defensive play, but it’s more important to find another gear going forward if we want to make a run at the playoffs like 22-23.

So what are our chances?

Given all of that, and where we’re starting from, what are Thistle’s chances?

I ran some numbers last year to simulate the playoffs, so I thought I’d give that another bash. I’ve tweaked the model to take more into account this time (I’m sure you don’t care about the detail). But let’s play out some scenarios and see where we might end up.

It’s worth saying that there are a lot of unknowns at this point - who’ll win the QF tonight, which Prem team will finish in 11th, and so on. So the model does this:

• Takes into account the actual results so far (i.e. the first leg of the QF last Tuesday)

• Simulates the remaining QF game based on the relative strengths of the two teams, and the home advantage, and recent form and attacking output

• If the simulation ends up level, play some extra time, and if it’s still level after that, simulate penalties (with a slight home advantage)

• Simulate the Semi-Final in the same way (based on who one the QF)

• Simulate the remaining games in the Prem to find which team faces the winner of the SF. This takes into account current standings

• Simluate the final in the same way as the QF and SF

• Repeat this many, many times (100,000 to be exact)

When you do that, Thistle win the playoffs around 38% of the time, and 67% of the finals we get to. Killie win another 31% of the sims, Dunfermline 22%, and so on.

Basically, at this stage, we win more than anyone else does, but that’s partly because we face one of three different Prem teams, so each of the Prem teams get fewer chances win (which they’re presumably happy with!). Dunfermline have almost the same win rate as us when they reach the final. I expect as the picture becomes clearer after tonight, and as the Prem settles to a fixed team in 11th, we’ll drop our win rate a bit.

But at this stage, I’m pretty happy with that.

I’ve broken down all of the scenarios in this last chart. We go out in the semi-final 44% of the time - in other words, we’re favourites even at this point to get to the final, but not overwhelmingly so. I’d like to think we’d be stronger favourites if we were the early season Jags, which we could still get back to, but I’d be cautious about declaring us sure and certain winners against tonight’s victors.

Our most likely route through is to beat Dunfermline and then Killie, both on aggregate (about 21% of sims). The only other scenario that happens in double digit percentages is losing to the Pars in the semis (27%). After that, we’re down to rarer combos (beating Pars, losing to Killie, both on aggregate - 9%, for instance).

At this point, it seems like the most likely path is that we’ll have to beat Dunfermline and Killie (which I wouldn’t mind, as it means we don’t have to drop Taylor and Smyth as we would against St. Mirren, or Ethan “The Major” Ingram as we would against Dundee).

It’s still a tough ask to get promoted this way. But Thistle are in the best position they’ve been in years. I can already feel the poisonous sting of hope seeping into my veins.

It’s time.

On the surface of it, it seems like a solid idea to promote teams developing talent, and not just shipping in players from elsewhere - at least if you’re ok with interventionist rule-making by the international governing body1. Press-ganging clubs into having to have a pathway into senior football for young players is something that (at least some) clubs in Scotland are - shall we say - in need of.

As Dougie Anderson pointed out on Twitter after a recent Herald interview with Matthew Anderson (a graduate of the Celtic B team, and currently with 6 assists and 2 goals in the Belgian Challenger Pro League), there isn’t much evidence of a good pathway for young Scots - certainly not at the Old Firm - and there hasn’t been for some time.

That being said, there is definitely a bit more evidence of the young Scots getting minutes in lower league loans (The Terrace is doing a fun series of posts right now on some of the best ones - such as this about Sandilands on loan from Hearts to Montrose, or Motherwell’s Olly Whyte on loan at Stenny).

But the general area got me thinking this week: what is the squad profile in the Championship?

FIFA’s proposal started that thought process for me by drawing my eye to ‘homegrown’ U21s (though exactly how that’s defined could be debated - do they have to be a native of the club’s country? A product of that club’s academy? Something else?). But I’ve thought for a while that there’s a correlation between clubs who have the right age profile and those who do well - not just ‘younger is better’ but the right blend of age and experience.

Time is our most precious resource

Let’s start by having a quick look at how teams allocate their minutes. If you lined up all of the players in each Championship squad from youngest to oldest, and added each player’s minutes to the total as you walked along the line, how ‘old’ would you have to get before you’d accounted for most of the team’s minutes?

Thistle’s minutes are relatively nicely distributed across the age ranges. Not many at the younger end, and quite a lot in the 24-26 bracket, but also not that many in the older 32+ range (compared to last year, this is a big shift).

Meanwhile, Dunfermline have a lot of young players gaining big minutes - their line climbs more steeply early in the ‘age lineup’ than anyone else, and Airdrie only catch up with them from 30 onwards. St. Johnstone are a bit slower to get going until you hit 24, then they accumulate minutes fast.

I’ve highlighted a band in the middle of 26-30, which is a rough guide for the ‘peak age bracket’ for most players. Obviously, that’s a bit subjective, but I’ll come back to it later.

To look at it another way - if you walk along that age line, how far along do you need to walk until you’ve added up at least 75% of the club’s league minutes?

Dunfermline hit that number at 26 - extraordinarily young. Airdrie are not much after them at 27. Thistle are the fourth oldest in the league - mostly because of the experience of our central defensive unit - at 31.

Ross County, Raith Rovers, and Arbroath all need to get to 32 or older before you hit 75%, which in and of itself is somewhat telling.

Brief diversion into Scots

What about Scottish players in particular? After all, that’s the FIFA target.

Under 13% of Thistle’s league minutes were played by Scottish U21s. In fact, almost no club got over 20% - again, Dunfermline are the protagonists here. However, both Thistle and Dunfermline cap out at about 70% of club minutes being played by Scottish players of any age. That’s still pretty respectable, but it’s interesting that not many clubs would hit the level that FIFA are aiming at.

So: who’s getting those minutes?

Squad profiles

Partick Thistle

I really like the visualisations that Opta Analyst produce showing where the minutes were going in each club, so I gave it a go for Championship clubs2.

Players are further up this list if they have registered more minutes this season. Further right means they are older. The colour coding indicates whether the players have a year or more to run on their contract, or if they’re expiring soon, or if they are on loan, or have left already.

A few things jumped out at me from this chart.

Firstly, Thistle have done a great job of getting more players who are coming into - or already in - that central column with peak age big minutes, along with some strategic experienced players. If we want to develop a player trading model, we want more people in that middle bar (I’ve broadened it out here to 24 to 30 rather than 26 to 30 that I had earlier).

Last season, I wrote about this topic, and noted that Falkirk had a good number of players in that peak age bracket, but Thistle had hardly any (mostly our squad was focused on players who were prospects or very experienced in the squad).

That seems to have changed. There are more dots in that central column or just about to enter it than anywhere else.

Secondly, the weighted age of the squad has come down a lot - under 27 years old. This is based on multiplying the player age by the minutes played, and then adding the numbers for every player together, and dividing by the total club minutes played.

Thirdly, there are a few players just about to move into the peak bracket in the next couple of years - including McPherson and Clarke coming back from loan on permanent deals - which again, will keep this profile strong in the next couple of years.

However, one concern here is the amount of yellow3, especially in the players further up. This profile could swing a lot of several of these players don’t extend their contracts next season. And having a very young squad is not necessarily a positive (see Airdrie for an example of that).

If you plot that average squad age weighted by minutes against each team’s points per game, there’s no massive ‘straight line’, though it’s clearly the case that when you average older than 28, the points per game drops off quite a lot. Similarly, skewing young makes it harder to gain a lot of points. You really need a balanced squad somewhere in the middle. You don’t guarantee a winning squad if you hit a magic number between 25 and 27, but you make it very hard to challenge for the league if your number isn’t in that regime. This is a small sample size, but it’s an interesting pattern. The goal is probably to peak in the middle of that curve if you want to challenge for the title.

What’s much more telling, though, is plotting the percentage of the squad’s minutes given peak age players against points per game. There’s no real overall correlation … other than how many minutes St. Johnstone gave to players in that peak range compared to everyone else.

Their squad was in that sweet spot; not just ‘guys who won the leage in the last ten years so might work’ (I’m looking at you, Dingwall), but guys coming into their best years right now. Guys who were one of the main performers at other teams in the league (Fowler, Boyes, Gullan, and so on), in their strongest years, with a sprinkling of young stars (Steward), and wizened heads (Holt).

Several of the players near the top of their age profile chart are still in red too - likely to stay at the club now they’ve been promoted. In particular, the peak age players (other than McPake) with 900+ mins all have a year or more to run on their contracts (Smith is drifting out of this zone). Above peak age, they’re mostly yellow - likely to be renewed if it makes sense, but the club isn’t tied in.

I think this is fascinating from a squad-building point of view. St. Johnstone’s profile of squad was clearly ‘win now’ (big minutes for those peak age players) but also ‘keep the assets you need’ … much more like Falkirk than Livi.

Meanwhile, Dunfermline are - as we’ve said - much younger in their profile. Most of their big-minutes guys are a) still under contract and b) under the ‘peak age’ window. They’re likely to get a lot better in the next season or two if they can keep the squad together (Matty Todd and JCM are the two guys they don’t have locked down yet). However, they have used a lot of players - 36 compared to Thistle’s 27 and St. Johnstone’s 29. They have money to burn and are using it on squad volume, which will help them as they come into the playoffs.

Arbroath are the other side of the same coin. Most of their big-minutes players are a) on expiring contracts and b) on the right-hand side of the peak window. This is their season to do something, otherwise it’s into aggressive rebuild territory pretty soon. If you lift loanees Marshall and MacIntyre out of that squad, and one or two of the big minutes guys leave (as Muirhead will), they’re going to have to work hard to maintain this season’s success.

The season is almost done - only tonight’s games left to go and then into the playoffs.

But what’s interesting to me is that, though age is just a number (after all, some of the most creative players this season in the league are young, like Marshall, Andy Tod, and Ben Stanway, or older like Dow at Arbroath), there is a definite sign that Thistle are moving in the right direction in terms of squad-building. Not just assembling a squad to challenge this season, but one which is designed to deliver on what was talked about in the close season - bringing down the squad age, and building a good player trading environment where young guys get minutes, but develop into compelling assets.

St. Johnstone have done it best this season. But whilst Dunfermline are backing youth and Arbroath experience, I’m encouraged that Thistle are on the right middle path.

This is a truth articulated by famous Thistle fan1 Fyodor Dostoevsky. Is it better to be a man, he asks, or a bird? Or, to put it another way, “If you can’t be good, be funny.”

This is the train of thought kicked off by a discussion in this week’s Draw, Lose or Draw pod - whether there was any replacement for a Banzo-booking-shaped hole in the hearts of Thistle fans.

That classic tweet still echoes in our hearts: Bannigan is booked.

So I started to look into what the shape of discipline was in the Championship over the last few seasons. Is the league more physical than it used to be? And does that physicality make for more discipline issues? And who at Firhill is picking up the slack now that we no longer have Banzo’s metronomic passing bookings?

First off, there’s an interesting trend here. If you look at the league as a whole since the start of the 21/22 season (i.e. when Thistle came back to the league after our sojourn to League One), the physicality of defending has gone down. The average player in the league gets into fewer defensive duels per game.

A defensive duel is any incident where a player physically engages an opponent to stop them from developing an attack. It might be a sliding tackle, or two-way tussle for a ball, or a through-ball cut out before it reaches its target. It’s a measure of how many times players out of possession get physically engaged.

This metric adds up all the defensive duels and all the minutes played by players and then converts it to a ‘per 90 minutes played’ number. So it’s a measure across all players who played (keepers, forwards, centre-backs and so on), and describes an ‘average’ player. Obviously, Ashcroft is more likely to get into a defensive duel than, say, Fitzy, and so the ‘average’ player doesn’t really exist. But it’s a useful measure of how physical the league is.

The peak point for the Championship in recent years was in 22/23, at 5.74 defensive duels per game. This season, we’re down to the lowest in five years, at 4.68. That’s 19% below the peak year, and 7% down on last season.

At the same time, though, the number of yellow cards per 90 minutes played has gone up noticeably - up to 0.19 from 0.13 in 21/22. That’s a nudge upwards on last season as well (0.18), and a big rise in a few seasons. Basically, on average, each team would get 2 yellow cards a game (ish) this year, up from 1.4 a game five years ago.

So the league is less physical but more likely to generate yellow cards. That could be a measure of the players in the league (for instance, they turn in fewer challenges, but they’re much more dangerous), or of the officials (more card-happy than they used to be), or of the rules (punishing things that used to be deemed OK), a combination thereof, or something else entirely.

But what about Thistle in that context?

Season by season, I’ve compared the rate at which the average league player generates defensive duels, fouls, and yellow cards, with the same numbers for an average Thistle player. Again, this is sort of a fudge, because there’s no such thing as average, but it’s a useful starting point.

Last season, we were above average for how much we engaged physically, we didn’t foul quite as much, but we generated more cards than the league average. The duel number is partly a feature of possession - we had more games where the opposition dominated the ball.

This season, it’s kind of flipped. Fewer duels, fewer cards.

You can convert this into a kind of ‘Criminal conversion rate’. How often do you turn a physical engagement (like a defensive duel) into a foul? And how often do those fouls turn into cards?

If you look at the % swing between those rates for Thistle vs. League average, you get an interesting picture. Most seasons (other than 23/24), we have a lower foul rate per duel than the league (the red bars are under zero), but a higher yellow card per foul rate than the league (yellow bars above zero).

Basically, most of the time, we don’t foul as often as most teams, but when we do … it’s a bad one.

The pinnacle of that was last season - we were 12% less likely to turn a duel into a foul, but 18% more likely to get booked when we did.

These two truths are somewhat correlated - if you don’t make tactical fouls as often, you are a bit more likely to misjudge it on the times you do. Though it’s also the case that when you get a reputation for fouling opponents, officials are more likely to call it bad at a lower bar.

However, this season, there’s been a bit of a tightening. We still foul less often, and convert those to yellows more often, than league average, but it’s much closer.

Is that the ‘Banzo-shaped hole’?

I’ve picked out players who played at least 900 mins this season and last, and grouped them by players who are no longer at the club, those who are new to the club, and those who are returning from last season.

If you add up the players who are no longer at Firhill this year - Banzo, plus Brian Graham, Kanayo Megwa, and Harry Harry Milne - that sums up to about 1 yellow card per 90 lost between them (50% of which comes from Banzo alone). Bannigan had a stellar season last year by his own standards. In fact, it was a top-10 return of yellows per game of any player in the last 5 seasons.

The returning Thistle players have basically been a wash this year in terms of yellow card production. I compared them to their previous ‘high water mark’ yellows per 90 season as a benchmark. Ashcroft, O’Reilly, and Stanway have gone up (in other words, are having a more yellow season than any previous Thistle year), but McBeth and Crawford have had a noticeable drop.

The new players have picked up some slack - mostly Smyth and Watt - adding some yellows to replace what we lost. But overall, the drop in yellows per game shakes out to about one yellow fewer every three games this year.

If you plot all the players who have played at least 900 league minutes for us in the last five years (66 of them), you can see the criminal conversion rate offenders.

It’s more normal to have attacking players further right - they generate fewer defensive duels per game, and so it’s easier to get a high conversion to fouls. Brian Graham has two seasons above 1 foul for every 5 duels, as well as more than 1 yellow per 5 fouls, and all four of the seasons he’s in the data he’s above 0.3 fouls per defensive duel.

Alex Samuel is pretty far right without getting any yellows - he picks up fouls, but generally not bad ones.

Lee Ashcroft, however, has picked up a high of over 1 yellow per 2 fouls this year - he doesn’t foul often, but he makes it count when he does!

Dan O’Reilly is a rare player this season - above 20% for both conversion rates. He’s not quite above 30% for turning fouls into cards - that was what made Banzo’s production so impressive last year - but he’s doing his darndest nonetheless (and Ben Stanway not too far behind!).

The last thing to touch on is player gravity. I’ve been working on this metric as a way of measuring players who are good at attracting opponents towards them. High-gravity players tend to draw one or more players from the other team towards them (or they drive into traffic themselves instinctively). Often, dribbly wingers score high on this - the single highest Thistle player here is Fitzy in the 23/24 season.

Some players generate yellows by being in the mix - by rushing in where others fear to tread. O’Reilly and Crawford are both pretty high-gravity players, often surrounded by opponents. And both have relatively high yellows per 90 numbers.

Interestingly, Smyth and Ashcroft are very low-gravity players (as Banzo was last year) - they tend to work their way into space, away from other players on the pitch, as much as possible. And yet, Ashcroft still manages a respectable return of yellows per 90, whereas Smyth has mostly kept himself out of the book for yellows.

Stanway, McBeth, and Reading have similar gravity to each other, but very different yellow card rates.

Basically - gravity isn’t a particular predictor of booking rates. But it is certainly more common for players who are higher gravity to end up with more cards (which is probably pretty obvious when you think about it! Finding yourself close to opponents is more likely to give you a chance to make a bad foul). 9 of the 13 players with more than 0.25 yellows per game were high gravity players. Only Crawford, Turner, and Banzo last season, plus Ashcroft this season, break that level as lower gravity players.

So: we’ve lost the King of the Cards this year. But whilst we’ve not replaced him - more redistributed than replaced - we’ve found our own way of being wrong this year. We’ve found other players who can get the cards in other ways.

If you can’t be good, at least be funny.

Or something.

One of the side effects of Chat(GPT)ting With Claude is the fact that it makes it possible to take on projects that you would never have dreamed of in the past, because it short-circuits the need to learn whole swathes of coding arcania (for instance) that previously would have thrown a 3-month roadblock in front of you (and therefore, to all intents and purposes, made the barrier to entry far too high).

I’ve been able to achieve side projects that would have been unattainable for me a few months ago. Which is - in equal measure - exhilarating, exhausting, and slightly terrifying.

As someone who’s spent a career thinking about gnarly data problems and seen what happens if you handle data too blithely, it makes me a little nervous that people can go from zero to ‘here’s my Advanced Scouting Metrics App’ in a few hours without any contact with either human synapse or real flesh-and-blood scouts in between. Some of the most valuable insights I’ve received offline have been from people who … y’know … watch a lot of football and understand what they’re watching.

But anyway, I digress.

One of the other benefits of AI removing some of the barriers to entry is the way it has allowed people who are really good at thinking about analysis and football a bit more capacity to think about it and offer insights. Not because they’re doing things that are more complicated. But because they have more brainspace to step back and look at patterns.

As a result, I’ve seen a bunch of really intriguing insights over the last few weeks that may not be directly AI-assisted, but are flowing with a remarkable fluidity in the context of an AI-greased world.

I’ve said before that one of the best ways I find to come up with ideas to write about comes from the best ideas that other people have, and I then work it out in our context - in the Scottish Championship, or for Thistle in particular. I am forever grateful for the way that Football Analytics Twitter is so willing to share an observation or an opinion from which I can bounce.

One such observation that has flowed down that stream of helpful insights this week was this Tweet:

You can go read the whole thread for yourself, but the gist is: the best, most scalable attacks in football have two different types of threat: threat that works looking in the eyes of the defensive line, and threat that gets behind the defensive line. And they have both of them at the same time.

Exceptional players can do both. But normally specialise in one way or another. And a balanced team attack needs both - not to be sucked towards one or the other (otherwise they become too easy to defend against).

Imagine the pitch above. The shaded area shows the average position of the defensive unit of an opponent. In this case, some attacking events happen inside that zone (the red ones), forcing the defence to move around, get drawn out of position, and open up gaps elsewhere.

Other attacking events happen behind the solid red line - deeper than the average position of the opposition defensive line - in a way that forces the defence to turn and track back, to play glancing over their shoulder.

Efficient attacks have both types of player - both players who get in about the defensive unit, in front of the ‘deepest defender’ and draw them out, or run at them, or pass through them and players who get behind the lines and force defenders to turn and cover.

It’s one of those observations that seems obvious. Of course, a team with more varied attacks will end up doing better going forward.

But there’s a lot of nuance in there (again, go read the thread). And it’s about squad-building and coaching, and tactical awareness, and teamwork from an attacking unit. How often have you seen a team with one attacking star who seems undroppable, but without whom the team play better, and are more balanced, in attack? Think Cristiano Ronaldo, one of the best players in a generation, struggling to lift his Portugal side who seem to play more naturally without him. Or, closer to home, wondering how we’d replace Brian Graham this year, and ending up with several players who each pick up some of the slack for a more exciting attacking unit.

Basically - an efficient attack needs both types of threat at the same time. Players who attack facing the defence, and players who attack by getting behind it.

So, I thought I’d investigate and see if this held true in the Scottish Championship.

What is a balanced attack?

First, let’s define terms. There are lots of ways to do this, but I’ve gone for these definitions:

An attacker who focuses on ‘in front of the lines’ is one who gets a high fraction of his touches and engagements inside the opposition defensive unit. Who’s not scared of engaging, but whose focus and bias is mixing it in with the defensive. Imagine the shaded shape on the pitch above showing the average location of the defensive unit. An ‘In Front’ attacker is in that shaded area a lot. It means that he’s not primarily going ‘beyond’ the defenders, but mostly facing up to them1.

An attacker who focuses on ‘behind the lines’ is one whose events mostly happen to the right of the red line in the pitch diagram above. Whose events occur deeper than the deepest lying defender.

And, for the purposes of this post, a balanced attack is when a team has at least two players who meet each of those definitions in the same game.

Well. That’s a fairly definitive chart.

Further right means a team meets the ‘balanced attack’ criteria in more of their games (2 in-front and 2 behind-the-lines attackers). And further up means the team is averaging more points per game.

This season, around 84% of the variability in points per game is explained by how often the team manages to have this sort of balanced, efficient attack.

St. Johnstone have managed it in about 90% of their games. Thistle in just under 80%. Then a gap to Arbroath and Dunfermline. Then another gap to Ayr and Raith Rovers.

It’s a bit more squirrelly when you get down to the weaker teams in the league. Ross County have managed to get a balanced attack around the same number of times as Ayr and Rovers, but don’t get nearly as many points per game (so there are other factors that this doesn’t account for, which is no real surprise). But above 65% or so of games, it’s pretty predictive of how well the team has performed.

But it’s very well correlated. In fact it’s close enough that I had to go back and double-check my code, because it looks suspiciously straight-line-y. Basically, the contention that the original poster had holds water. The more effectively you get two attackers doing each job in the same game, the higher your points per game.

Who, then, provides that threat?

Individual contributions

Let’s look for players who do one or other of the types of threat in a lot of his games. I’ve limited it to games where players had at least 30 events, just to remove some noise. And Thistle players are highlighted (in a + in red), as are St. Johnstone players (diamonds).

Some players provide threat both ways - in front of the lines (further right means closer to 100% of games where they meet the threshold) and behind the enemy lines (further up). To the right of the vertical line means the player hits the mark for ‘in front of the line attacker’ in at least half their games. Above the horizontal line, and they hit the mark for ‘behind the lines’ in at least half of their games.

There are a few freakish outliers. Curtis Main at Morton was an ‘in front’ attacker in every game. Josh McPake (more on this in a moment) and Callum Morrison both get behind the lines in 100% of their games. Which is pretty extraordinary.

These players provide one type of threat and do it exceptionally well. But if that was all their teams had, then they would struggle to have an efficient attack.

Main doesn’t have as many supporting players backing him up in front of the lines, and their behind-the-lines threats don’t do it often enough to help him out. Morrison is playing in a side marked by injuries to attackers, making it hard to get this sort of impact at a team level - and he ends up dropping behind the lines more than McPake to do both jobs as a result.

McPake, on the other hand, has help - Sidibeh backed him up behind the lines up to January, and Fowler, Gullan, and Mallan can all chip in here too. In front of the lines, McPake doesn’t have to do much because Holt, McAlear, and Mallan are all pretty potent here (and Gullan and to some extent Fowler can moolight here if needed).

It’s interesting that many of the top ‘behind the lines’ players in the championship right now are wingers (Fitzy and Chalmers both do well on this metric too, and the fact that we’ve often been best when both of them can do it means teams can’t focus on one flank). And Thistle have two two-way threats who can contribute either role given a particular set of instructions - both Watt and Samuel are above 50% for both types of role this season, which is really powerful for our attack. Depending on the game, and the team, we can use them in either ‘mode’.

Several other Thistle players are at or near the threshold to contribute, which means that we can rotate the attack when we need to.

If you look at specific games, the top 20 games for ‘in front of the lines’ attack features three Thistle performances - two from Tony Watt and one from Robbie ‘Bob’ Crawford. I’ve only included the players who are not the high-possession midfielders in this list (as the top 20 are otherwise almost all Holt and McAlear at St. Johnstone and Smyth at Thistle. These players get more chance to touch the ball, so end up near the top of the tree for this metric). Again, it’s interesting that several Ayr players feature in this list - but Ayr’s problem has been an inefficient attack. Even in a game where McKenzie, or Banzo, or Dowds, or someone else, poses a lot of threat in front of the defence, they haven’t managed to get the runners behind the lines to balance the attack out.

It’s also worth keeping an eye on Innes Cameron of Raith Rovers, who makes the top 20 here - he’s had a very fruitful few weeks, and pairing him with Dylan Easton helps their attack a lot.

But the really striking chart is this one. This is the top 20 games by number of events behind enemy lines. 12 of the top 12 games this season have been Josh McPake. Then a brief diversion for Fitzy and Watt, and then back to McPake. In total, 15 of the top 20 games this season have been McPake.

It’s much easier to build a balanced, efficient, attack if you have a player who you know will do one part every single game the way McPake does.

He’s had an extraordinary season, and this is one of the big reasons why St. Johnstone’s attack has been so impressive (and individually why Hearts have snapped him up). When you pair him up with Gullan, or Fowler, and then have McAlear, Stanton, Holt, Mallan, providing threat from in front of the defence, it’s pretty hard to resist them.

All of this is a long way to say: I think the original poster was right. In our league, having a balanced, efficient attack goes a long way (84% of the way) to getting you a great points haul. So recruitment and squad building isn’t just about getting one standout player (though if his name’s McPake, that probably does help).

It’s about building balance.

I’m off to the game today for the first time in a while, and I can’t wait. Who doesn’t love the smell of a freshly-baked Munns, or the feel of a Jaconelli’s chip between your toes? Wait. No. That’s not quite right. You get the idea.

Anyway, I don’t get to the ground often these days, living as I do further away from the vaunted tunnels of Firhill than facilitates a quick jaunt each Saturday Friday night. But thanks to the generosity of Mrs Jagalytics, left in charge of the home fires and the Junior Jagalyticians, for a few days, I’m looking forward to the opportunity to watch the team live.

So this won’t be a long post. But I’ve built three charts on the theme of “Opportunity”. Some players score a lot of goals mostly because they take a lot of shots. Some keepers make a load of saves because their defence is really poor at protecting them.

But which players make the most of every opportunity that presents itself to them?

Shooters

First off: in front of goal.

Further right on this chart means the average chance players get is better. Not that the shot itself is necessarily good or bad, but they consistently find shots that have a higher likelihood of going in. You’d expect to find Brian Graham at the far right of this, for example.

Players further right give themselves more opportunity to score, because the chances are better to start with.

But further up on the chart means the shots are improved by the player. If you’re above the ‘zero’ line, the way a player takes the shot make the chance better after it leaves his boot (/head /backside). In other words, the player’s shooting technique is good.

It’s harder to be much higher up the further right you are. It’s harder to make a 0.2 xG chance into a 0.3xG chance than it is to make a 0.1 xG chance into a 0.15 xG chance, even though both are a ‘50% improvement’. So it’s not surprising that the goal-lurkers like Fowler don’t often break over the horizontal ‘zero’ line.

In fact, only six players in the league average chances that are pretty good to start with (0.14 xG and up) and whose technique improves those opportunities. And two of them play for Thistle (Samuel and the hopefully soon-to-return Tony Watt). That’s both a measure of the player’s technique, but also, I suspect, of good coaching to encourage already-good players into good positions.

Disruptors

The second ‘opportunity’ chart describes who makes the most of being in and out of possession.

Further left on this chart is ‘good’ here - for every 100 touches a player has, how often do they lose control of the ball. On the whole, poachers will do badly here - players who play further forward are more likely to give up the ball a high fraction. Anything under 20 is very strong - in other words, left of the 20 line means the player keeps control of possesion with 80% or more of their touches. Above 30 tends to be the more wasteful players.

The other axis - bottom to top - is bit more nuanced. Out of possession, how well do they disrupt the opposition and regain the ball for their team? The actual number isn’t that important, but the idea is - for every 100 or so times the other team touches the ball, how often does the player intercept passes and regain possession?

So: further left but down low are players who are good on the ball but less likely to regain possession (for instance, Josh Clarke is great here - he hardly gives the ball to the opposition).

Further right, but high up, are players who are a bit more wasteful in possession, but who regularly flip the ball back to their team when the opposition are building up. Guys like Ross Docherty live here, as does Aaron Muirhead.

The top left, though, is where you really want to be.

Nurudeen Abdulai of this afternoon’s opponents is prominent here. Jeremiah Chilokoah-Mullen doesn’t quite get below the 20 line, but is good at intercepting a lot. The fact that both of them are around this area tells us something we’ll need to track today - they will snaffle up any loose passing as we go forward, so our ball use needs to be controlled, tight, and disciplined.

Our defensive unit is very impressive though. O’Reilly and Smyth are in the top left zone very comfortably, and Ashcroft (who of course won’t be playing today) is there too (though a little bit looser and a touch less likely to intercept). Paddy Reading and Luke McBeth are both very close to the golden zone (just a touch more wasteful than the 20 line), and McPherson, being a bit more attack-focused, doesn’t pick up as many interceptions.

But as has been the case throughout the season, though we have outstanding attackers who get a lot of credit, the bedrock of Thistle’s success is the defence.

Creating

The last chart is about attacking creativity. Who makes the most of the chance to create a chance?

Further right on this chart means a player is more likely to turn every touch into a dribble attempt, taking on his man. Obviously, Josh McPake is in a league of his own this year in this. More than 16% of his touches turn into an attempt to beat a player 1v1.

Further up means a player needs fewer passes to create good shooting opportunities for team-mates - either because their passes create a lot of shots, or the shots they do create are really high-value ones.

As I’ve said before, Fitzy is having a stellar season in this second metric. His passing is deadly. Only two players in the league generate more expected assists from passing than him.

And overall, only (soon to return) Logan Chalmers and one other player join Josh McPake in the ‘top right’ zone - creating more than 0.7 xA per 100 passes, and turning more than 10% of touches into take-on attempts.

One interesting side-bar of this chart is that Morton have the second-best dribble-instinct player (Garrity is surprisingly close to McPake), and the best xA-generator (Nathan Shaw). Who’d’ve thunk it?

Bonus chart time!

OK, I lied. I know, I know, I’m sorry. I don’t know why, but I’m fascinated by players and teams who get goals from throw-ins. Did you know that last season, Jaze Kabia (once of Livi, then of Truro, now of Jamestown-fuelled Grimsby) scored seven goals that started life as throw-ins in the National League North? Kooky.

Anyway, who gets the most opportunity to score from throw-ins?

It’s Findlay Marshall. Maybe the wind at Gayfield helps. But he’s scored two goals this year (one of only three players in the league who’s scored more than one this way) from set-piece routines starting from the sidelines, and he’s done that from just over 1.3 post-shot xG. And look at how many more shots he’s taken in this sort of situation than anyone else - 14! That’s exactly twice as many as Brophy and McKenzie, who are next highest.

Meanwhile, Adama Sidibeh, with only half a season, took five shots after throw-ins, and scored three of them, which is pretty nifty too. There are actually a few goals missing from this list, for boring data capture reasons (for instance, Thistle actually scored after an Ingram throw-in recently against Ross County). But the general pattern is: Findlay Marshall is the Championship’s Mr Throw-In Receiver.

Overall, from the first three charts, it’s striking that the ‘opportunity players’ from Thistle who feature strongly include those that are hopefully back from longer injuries in the next game or two - Chalmers for creativity, Smyth for disruption and possession, and Watt for shooting.

Right. That’s it for today. I’m looking forward to the opportunity to watch the team run out in person today. Let’s hope I get to make the most of that opportunity.

And a race it is.

One would have been forgiven for thinking that two weeks ago - with the spine of the Thistle team out for several weeks through international call-up, suspension, or injury - it was just about digging in. Not deflating too much. Clinging on to what we had.

But Mark Wilson had other ideas. Without the core of his team, and looking down the barrel of a potential double-digit gap to the league leaders, Thistle drew with St. Johnstone, and then (converting Ben McPherson to Logan Chalmers 2.0, and watching Robbie ‘Bob’ Crawford1 break his league duck in the process) broke the Starks Park no-win streak and put Ross County to bed. Meanwhile, St. Johnstone didn’t quite manage to overcome Queen’s Park. And the gap was down to 3 points plus goal difference. Game on.

Wilson, in his Jagzone interview today (clips here), called for consistency from his players as we go into a crucial weekend in the title run-in.

Consistency. In many ways, that’s been the watchword of Wilson’s tenure. Keep the heid. Stay the course. Don’t abandon the plan when there’s a wrinkle or a bump.

One of the most impressive features of the Raith Rovers and Ross County games was the consistency with early-season style, despite the missing core.

I wrote last year about one way players can be consistent (I’ll return to this in due course, though not today). But I thought it might be interesting this week to look at some numbers to find who in the Championship is truly consistent - who keeps the head when all around them is losing theirs, et cetera, et cetera.

Team level consistency

First, let’s set the scene. There are lots of ways to quantify what teams are ‘like’ in and out of possession. But let’s keep it relatively simple. In the chart above, further right means the ball pings around more when a team is in possession. Further up means they’re more likely to stand off when the opposition has the ball.

Thistle are the third fastest at moving the ball around in the league. And the fourth most intense pressers - teams get far less time on the ball before we get involved.

But does that vary from game to game? Our results are notoriously different away from Firhill this season. But do we play differently?

This chart shows the same two metrics, but measures how consistent teams are from game to game. It looks at the variability in passing tempo and pressing intensity and compares each team to the league average. Above 100 means a team is more consistent than the league-wide average. Which is to say, more likely to ‘stick to the plan’ and play the same way game after game.

Teams can be good and consistent … or bad and consistent. But at least if you know your target plan works, being able to recreate it is very helpful.

Thistle are about in line with league average in terms of in-possession. We’re pretty consistent, but a bit more variable than, say, St. Johnstone or Dunfermline.

But out of possession - how much we make life difficult for opponents, we’re one of the two most consistent teams in the league. Which rings true - with the experience and performance of O’Reilly and Ashcroft, the explosive impact of McPherson, and Mr Dependable Paddy Reading, the defensive unit has clearly been a picture of metronomic consistency this year.

So: we’ve got some features of consistency. What about individual players?

Overall

Of course, there are lots of ways to measure overall performance. So let’s just pick one: the ‘Sofascore Rating.’ On the Sofascore app, players are given a score from 0 to 10 (realistically, normally 6 to 8) using a proprietary algorithm2 for each game they play. Their position is taken into account, as is their overall contribution.

Higher scores are harder to keep consistent. If you get a few 10s, and a few 6s, that might average out to a 7 or 8, but there is much more bouncing around. Explosively good players (like Josh McPake) who sometimes blow opponents away will be a bit more likely to have variation because you can’t get a 10 every game.

But of the players with 900+ minutes played this season, it’s worth homing in on the players who get an average score above the league ‘normal’ of 6.9, and are more consistent than their peers. Dylan Easton, John Tod, Jason Holt, and Lee Ashcroft fare well here. They all average above 7.2 per game they’ve played, and don’t bounce around much. They consistently get that score, come rain or shine.

A lot of these players are defenders - it’s harder to be consistent and good if you play up front or in midfield. But Holt, as well as McAlear, Euan Henderson, Easton, and Jamie Murphy all perform pretty well.

Thistle’s most performant players in this respect are our Captain and Vice Captain - which is probably a Pretty Good Thing.

Passing

Let’s zoom in a little bit to one particular feature - passing the ball.

This time, I’ve only plotted players with at least 600 passes for their team over the season so far (which rules out several goalkeepers). What we’re interested in is who is most reliable? Who gets a high average of their passes to a team-mate? This is slightly biased towards players who make short and safe passes, but even so, if you want to use the ball well as a team, it’s good to get a high percentage of your passes meeting their mark over and over again.

The best players in the league tend to hit at least 75% per game. Six players average north of 85% per game all season - four St. Johnstone players (Holt, McAlear, Steward, and Diabaté), plus Charlie Gilmour and our very own Ts’oanelo Lets’osa.

All six of them are very consistent (which is not that surprising, as to get that high an average success rate, you need to do it game in, game out).

So Lets’osa is one of the most consistent passers in the league. O’Reilly is above 80% accuracy, and over 200 in the consistency rating (and Ashcroft not far behind). Clarke is the only other keeper in the league above 80% accuracy and north of 100 on the consistency axis.

It’s interesting to me that - once again - our consistency is driven by our central defensive unit. But it’s also telling that St. Johnstone dominate this chart. Three of three in the top right corner are their spine, in the centre of the park. Holt, McAlear, and Diabaté are crucial to how they play, and are the pulse of the team, making sure the ball gets where it is intended to go over and over again.

Lets’osa is fascinating to me on this chart. He doesn’t often get the plaudits as a crucial part of our team. And yet, he is clearly the most reliable, consistent passer in our side.

Duelling

The final metric I’ll touch on today is the more physical side of the game.

How consistent players are at the more engaged side of the game - battling in contested situations with opponents to win the ball. The term ‘duel’ basically means any situation where two or more players are fighting over possession - either tussling on the ball, or jumping to win a header, or running after a loose ball. It can be in an offensive situation (where a player is taking on his man in a 1v1 dribble), or defensively (last man against his opponent).

Who consistently comes out of those situations with the ball?

This chart only has those who have been in this situation at least 100 times over the season. These players, on average, come out with the ball just under 40% of the time. Only two players on the chart win more than 55% of their duels - both of them Dunfermline defenders (who’d’ve thunk it?). Tod and Benedictus are both clearly the most consistent players, always winning more than half their duels in every game.

Again, the top right of the chart is generally dominated by defenders - O’Reilly and Ashcroft are in a little picket who are around 50% more consistent than league average, and tend to win 50-55% of their duels.

Most of the out-and-out forwards in the chart are in the bottom left - they tend to win under 40% of their duels (other than Jamie Gullan and Tony Watt), and are fairly inconsistent - they’ll have good games and bad games. Games where they win lots of contested balls and ones where they lose most of them. Alex Samuel is more consistent than average, and tends to win about 1 in 3 of his duels.

The three exceptions to the ‘mostly defenders’ principle are Ross Docherty (wins about 50%, and is about 50% more consistent than league average), and then the St. Johnstone pair of Taylor Steven (though when he was at Arbroath), or Josh McPake, whose duelling has been one of the highlights of his output this season.

So: what did we learn?

As a team, Thistle are one of the most consistent out-of-possession teams in the league. And at an individual level, much of our consistency rests on the team leaders - Ashcroft and O’Reilly are consistent passers and duellers, and their ratings are consistently high.

We have a few other individuals who do well time and time again - Lets’osa is one of the most consistent passers in the league, and Tony Watt one of the most consistently effective attacking duellers.

But the most telling feature is that St. Johnstone dominate many of these charts. They have the most consistently accurate passers (in Holt, McAlear, and Diabaté), and some of the most consistently effective duellers. Several of their players remain consistently above league average ratings week in, week out.

The question at hand, is can they do it for the remaining five games? Or will Thistle’s recent surge be enough to sneak past them at the last minute?

I’m so excited that we’re into April and still in a title race. Let’s find out how the narrative plays out. Wilson's called for consistency. Bring it on.