Get Your Stats Right: September 2024

The second monthly round-up of the numbers behind the U's performances

It’s an absolute pleasure to have the brilliant Ben Griffis writing these new monthly columns for UTAS, delving into the numbers that make up the U’s performances using his own statistical data and modelling. I’m sure you’ll agree this is a fascinating start with just a small sample size, made all the more impressive by the fact Ben is a US-based U’s fan. We’re massive now, it seems.

Give Ben a follow to discover more of his brilliant work, and subscribe below to get the third instalment directly to your inbox.

---

Another month of the season draws to a close, so welcome to another edition of Get Your Stats Right (GYSR, pronounced “geeza”). While we do have one final game this month playing host to Lincoln, at this moment we are still winless in the league with just 1 point from 6 games.

I was relatively buoyant in last month’s edition, illustrating the relatively decent underlying numbers that United had put up over the first 3 games. The results were not there, but the general early-days indicators, while not enough to make conclusions off of, signalled that if the club kept performing in that manner the points would come.

After 6 games, however, I will say that I’m a little less enthusiastic. Some of the key numbers I like to look at to help me understand a team’s “table health” are iffy to say the least.

I discussed field tilt in last month’s edition, and noted how it was possibly the one underlying piece of data I was most worried about. And after this month, I’m still worried about it. Field tilt is the share of a team’s possession in the final third of the pitch, with a lower field tilt indicating opponents have more touches near our box than we have near theirs.

After 3 games we ranked #23 for field tilt. This month we rank #24. Not much of a difference but it’s not a good sign that we’re seeing no improvement so far. Then when we compare field tilt to possession, we see that United rank #22, the same as after 3 games. This -8.08 number means that overall, if we have 50% possession we’d expect 42% field tilt, and is an indicator of the U’s passing the ball around deep, rather than being able to get it up the pitch consistently.

Why is that something I worry about? Well, we all see this on the pitch each weekend. United are able to retain possession better than in recent seasons, but that possession is usually with defenders and/or in our own half. It’s more difficult to create scoring chances when we can’t put the pressure on opponents consistently. And perhaps worst of all for this season…

Deep possession magnifies & exacerbates defensive lapses, mistakes, and errors.

Keeping the ball relatively low typically leads to opponents starting their possessions closer to our goal, which means there’s less work for them to do to get into the box and score. Mistakes from players in possession turn the ball over in relatively more dangerous areas.

While keeping the ball can be a decent way to defend, usually this tactic is employed by higher possession teams (United average 48% possession, not the 55%+ that’s typically needed for this tactic to be effective). Further, this style is mainly employed by teams who have possession much higher up the pitch, where losing the ball is relatively less dangerous and it’s often easier to counter-press because players are closer together. So, we’re not necessarily keeping the ball deep to defend, we’re just finding it difficult to march up the pitch consistently.

And this all leads to my next point: Under-Performance.

The U’s have a goal difference about 4 goals worse than expected this season, ranking 21^st. This comes out to an underperformance of about 0.68 goals per game. That could come from conceding more than expected, or scoring fewer than expected, or both. In United’s case, it’s sadly both.

Given this information, we can see how having the worst field tilt in the league (and beyond that, some of the deepest possession) causes the club loads of issues and exacerbates any problems we have. Deep possession, in United’s tactic so far, leads to fewer chances. And the attackers are scoring fewer goals than expected given the quality of their chances. Deeper possession can also make it relatively easier for the opponent to create chances. And we’re conceding more goals than expected, given the quality of the chances we’ve conceded.

This brings me to another foundational component in football analysis: Game State.

Game state indicates whether a team is winning, drawing, or losing at any point in a match. It is a simple concept, and one of the most important context variables for football data that is sadly ignored by most. The reason why game state is so important is that teams almost always change how they play depending on if they’re winning, drawing, or losing. These changes can be big or small. Sometimes the changes are a tactical instruction from the manager, other times it’s a mental component, and often it’s a combination of both.

For example, if a team is losing by one goal (a losing game state), chances are they will take more shots than if they were leading by one goal. And the quality of these shots, the xG of them, is likely to be lower than the average xG of shots taken when leading. Teams often generate more xG when losing than when they are drawing or winning. Even if the average xG per shot is low, they tend to take more shots in general. And this makes sense. When a team is losing, they’re more motivated to try to score than when they’re ahead.

On the other side, teams in the lead tend to sit back a little more to protect their lead. They become a little more risk-averse and don’t try to force a goal if it will leave them more open at the back. We can see how this is a feedback loop as well, as teams who are leading will usually be less risky and shoot less, and teams who are behind typically riskier and shoot more.

The third game state is Drawing. We typically see the best representation of a team’s tactics and true performances when they are level. As there’s equal risk of going behind or going ahead, teams usually set up in their most balanced approach.

Now, how does this relate to United? Well, to bring the worst news first… The U’s have led for 0 minutes across all 6 League One games. We’re winless and have conceded the first goal in every game so far. Our 1 league draw, that crazy 4-4 with Blackpool, obviously saw us come from behind to rescue a draw.

There’s only one other winless team in the league. Burton Albion. But the Brewers have been in a winning game state for 42 total minutes spread across 3 games (a loss to Lincoln, and draws with Rotherham and Mansfield). That means United are the only team in the league who have never been in the lead. A sad reality!

Now let’s look at some basic underlying numbers split by game state, and how they stack up to the rest of the league.

When drawing, the U’s rank 22^nd for xGD/90, or expected goal difference per 90 minutes (excluding penalties). Our number of -1.01 means that we could expect to lose by 1 goal in a hypothetical world where the game state is always even. Given it’s almost exactly -1, I’ll joke that if we played 90 minutes at 0-0, the last kick of the game would be an opposing goal. Only Shrewsbury and Bristol Rovers have a worse xGD/90 when drawing.

But how do the actual goals scored and conceded compare to our expected numbers? Not great.

United have recorded the worst performance vs their expected performance when drawing. A GD-xGD/90 (goal difference minus expected goal difference per 90 minutes) number of -1.81 means that not only are we expected to perform poorly when drawing, but we’re also going below and beyond that to perform even worse than expected.

While drawing, we have 3.09 xG and 0 goals. We also have 6.32 xGA and 9 goals conceded. The team is not making it easy to pick up any points at all, because we are always conceding first and then forcing ourselves to get back in the game, and if we do get back in the game, we’re still pretty poor.

But how are our performances after we’ve gone behind? When we’re in a losing game state? Thankfully, a little better.

Our xGD/90 when losing is 0.09. That’s close enough to 0 that we can say that after we’ve gone behind, we might be expected to keep the score as is, in some hypothetical world where we’re always trailing (is it that hypothetical though).

We’ve generated 5.13 xG and conceded 4.85 xGA in the roughly 300 minutes we’ve trailed in games. Compare this to the 3.09 xG and 6.32 xGA in about 300 minutes of drawing and we get information to support what I mentioned earlier: teams usually sit back a little bit more when leading, allowing opponents to generate more xG, and teams losing typically take more risks going forward to get shots off. We have more xG and less xGA when losing compared to drawing, which supports that notion.

The graph below shows how, beyond having a decent xGD when losing, we’ve also slightly over-performed as well.

We rank 5^th for GD-xGD/90 over-performance when losing. While this looks really good, I have to caveat it by saying 4 of our 6 goals from losing positions came against Blackpool. While that game shows the mentality the players can have when looking to rescue points, it being 1 of 6 games so far does skew the numbers slightly in our favor. Basically, we can’t be expected to always come back like that.

But, as I said, that game does show the potential the club has to improve results. We CAN turn losses into draws. Even just getting a string of 2-3 draws in a row would be welcome at this point, and allow the team to feel some confidence in the system that could help things click, improve our performances when drawing, and get some wins.

Wrapping up: team profile

All of these numbers (and of course the performances on the pitch, which are also what generate these numbers) tell us that something needs to change. The tactic, perhaps not in theory but certainly in practice across these 6 games, is only hurting us and it looks to be leading us towards League Two. Not nearly good enough when the game is within our reach, although the saving grace is we’re not as woeful when we go behind. The issue is just that we’re bad enough that we always go behind in every game.

But, there’s still 40 games left which is more than enough time to get out of a relegation battle let alone just not go down. But the style may need to become a little more direct just to get some results under our belt and then in January hope to bring in some improvements that further allow Monk to play how he wants – and for that style to then be effective on the pitch.

Now, with 6 games under our belt, I feel a little more comfortable introducing our “team profile” radar. This is something I’ve created that is based off of Analytics FC’s “Coaching Profiles”. It compares a team’s style in key style indicators (not “good” vs “bad” performance ones like xG) to others around the world. I have over 1,000 teams loaded at the moment. Overall, my metrics follow the same definition as those in the Analytics FC article I link above if you want more info.

The numbers are all percentile rankings, which are between 0 and 100, with 0 being the team with the lowest number in a metric, and 100 being the team with the highest number. So a team with a “Long Balls” percentile value of 95 can be said to play long balls more frequently than 95% of all teams.

Here is Cambridge United’s profile for these 6 games so far.

Please note that there’s still relatively little data, so these numbers are going to change, but likely not by crazy amounts since after 4-5 games, most teams solidify their style.

What does this tell us? I think it illustrates our play relatively well. Going clockwise from the top:

United defend in a relatively low block, 65^th percentile. This means we’re not really high up the pitch, but also not completely parking the bus. We defend closer to our own goal than average, though. If you’ve made it this far in the article, hopefully you can see how our low field tilt may play a role in this!

We have a pretty average high press, which is PPDA (passes allowed per defensive action). Monk isn’t press-averse, but isn’t instructing players to be Klopp’s Liverpool either.

United are not a counter-attacking side. The calculation for determining this metric is the percentage of open-play shots taken on the counter. Opta has tagged 0 for us, which means that when we counter we are not able to take shots. So we don’t pose much of a threat there.

We cross a little more than average, ranking 62^nd percentile. This is calculated as how United enter the box. We enter the box with crosses rather than passes a little more frequently than other teams.

Wing play is a little more complex, but essentially, it’s how we “escape the buildup zone”. United are typically moving possession laterally or vertically up the middle (or carrying) rather than passing onto the flanks. Play on and down the wing typically comes further up the pitch, not in buildup which is what this looks at.

Territory is just field tilt, so it’s not a surprise to see United ranking 7^th percentile! This means that of the 1,040 teams loaded in my sample, only 7% of them have a lower field tilt than United do so far. As I’ve been stressing, this is not a very good sign when taking into account our total performances.

Circulation is effectively lateral passing in deeper areas. While we do pass laterally here, there’s also a bit of verticality. We also pass it so deep at times that the lateral passes between the likes of Rossi, Digby, Andrew, and Reyes aren’t even captured in this.

United do involve the goalkeeper Reyes in buildup a bit more than most teams, but this is partially down to the fact that we have possession so deep that Reyes is much closer to the ball than other teams. Many teams with high GK Buildup numbers tend to have possession slightly higher up, and push the GK out of the box a bit more. Regardless, United still use the GK more often in possession than many teams. And this again harks back to what I’ve mentioned a few times now, that the team’s deep possession is magnifying and exacerbating the team’s poor defensive record.

Finally, long balls. We are not pumping the ball up the pitch. Passes are typically shorter and on the ground, not direct route-one “Brexit Ball” stuff. And to be brutally honest, I think this is in fact the direction the club needs to stay in. We need to add more directness to our attacks, but they should be built on a foundation of decent possession, placing opponents where we want them with shorter passes before switching gears with a long ball.

Hopefully, this graphic and discussion makes sense. I’ve also got code to find the most similar style teams in the sample on the bottom right. Norwich City from last season rank as the 2^nd most-similar style to United so far, and their graphic is below.

We can see the similarities, particularly in the high use of GK in buildup, a relatively low block and reliance on crossing, and very low territory/field tilt, wing play, long balls, and counters. Norwich finished 6^th in the Championship last season so hopefully this comparison offers a little bit of hope at the end of a long article where I’m airing my concerns!

Finally, compare our style this season to last season. Much more use of the GK in buildup, much lower reliance on long balls, and the team presses a bit more than it has. Seeing drastic style differences in foundational tactical elements like buildup, long balls, and pressing (which we can see on the pitch too) also helps explain why the players may be finding it slightly difficult to get Monk’s tactics to fully click. It’s very difficult to get players used to playing a style that’s almost the opposite of the style they’ve been used to playing.

Getting a host of new players in was important, but there’s still going to be a learning process for everyone. The results have been poor so far, the tactics have not clicked at all yet, and I think the tactics need some tweaking, but Monk deserves time to get it right. If we’re on 3 or fewer points after 10 games, however, that may be a sign that things won’t click, or when they do it’ll be too late. Time will tell, but right now, from the numbers, we’ve got some negatives to dwell on and improve while sitting on a couple positive elements.