Get Your Stats Right: December 2024

Just when you thought 2024 was done, here's one last stats round-up to see out the year

Well, I hope everyone had a wonderful holiday period with friends and family, because the 5 games from December 20th through to January 4^th saw 5 straight losses where the U’s scored just 2 and conceded 12. Scrooge must have been a Cambridge United fan.

We are now just past the halfway mark of the season. 24 games played, a record that reads 4 wins, 5 draws and 15 losses, and we sit 23^rd in the league, 8 points behind Bristol Rovers & Peterborough in 20^th and 19^th respectively. Not exactly where we want to be, and it looks even worse when we remember that 3 of those 4 wins came in a row in October.

United had a decent 10-game record from mid-October through to mid-December, with 4 wins, 4 draws, and just 2 losses. 1.6 points per game is top-half form. Over the past several years, that form over a season would land a team around 8^th, and Leyton Orient – who are in 8^th right now – have averaged 1.55 per game.

Monk and the squad did a decent job in this span to chip away at the mountainous points deficit that our 0-1-8 record in the first 9 games gave us. But sadly, this new 5-loss streak has undone much of the gap-closing.

Underperformance Trend

It’s hard to make many conclusions on trends after even 15 games of a season. 20 is a better number, and now after 25 I’m ready to make a sad conclusion that we’ve all known for a while. But I wasn’t comfortable signing off on the data to back it up.

This team’s results consistently underperform what “should” be the result based on the on-pitch performance. Week in, week out.

I won’t go into too much depth as I covered this in last month’s edition, but the main culprit is defence. According to FotMob, we’ve conceded almost 13 more goals than expected, and scored about 5 fewer. I noted that a lot of this, in my opinion, could be down to Vicente Reyes’ relatively poor shot-stopping form this season. Reyes should have conceded about 27.2 goals, according to FotMob’s post-shot expected goal model, but has conceded 36.

Reyes has been benched for the past 3 games, with Jack Stevens getting the starts. In this span, according to FotMob, Stevens should have conceded 5.43 and conceded 6. Including his game vs Barnsley, he’s faced 6.13 psxG (post-shot expected goals) and conceded 7 goals. Based on this, Stevens appears to be the better option for bailing out the defence.

Opta places us at 15^th on the expected table, which should help illustrate how much (and how consistently) we’ve under-performed across this season. We have 15 points fewer than expected, which is by far the worst in the league. Burton and Peterborough both have 8 fewer than expected, so we’re almost twice as bad as the second-worst!

Improve Attack or Defence?

Now that it’s January and the window is open, the big question is should we focus on attack or defence? Naturally, we need to stop conceding so many more goals than expected. I’m somewhat hopeful that Stevens can help with this. Stevens has let in about 0.22 more goals than expected over 4 games, and Reyes has let in about 0.44 more than expected. So theoretically, we could be half as bad defensively if Stevens stays in this same form over the second half of the season. That would really do us a world of good to help turn some draws into wins, or losses into draws without any other changes in the squad.

Below is a great visual showing how a good xGA is typically better than a high xG for relegation battles from Soccerment. While this has data on Serie A, Ligue 1, Bundesliga, La Liga, and the Premier League, we can probably assume it’ll be similar in League One. This data is why I think continuing to play Stevens if he keeps this form up (or even improves) would be better than playing Reyes. Our xGA is quite good. We’re just so low on the table in large part due to the goals we’ve conceded above expected.

Given the context that we have a decent xGA record but are under-performing (with the initial showings of improvement there), I decided to build a quick simulation to see if we should focus efforts on further lowering xGA or on improving our xG.

The expected points model in my app is fairly close to Opta’s expected points, and is also very close in terms of expected positions for all teams. It is built using a Pythagorean Expectation equation, which you can read more about here. In short, it divides a team’s xG by their xG + xGA, with xG and xGA being raised to a constant power, 1.54. This equation is shown below.

Using this equation, we can plot xG and xGA per game and shade the graph by the expected points (over a 46-game season) based on where a team’s average xG & xGA per game lie.

The lines on the graph below show Cambridge United’s xG of 1.14 and xGA of 1.26 per game, per Opta.

One of the things this graph shows is that there’s many ways to get the same expected points. Low xG but low xGA could be the same as high xG but high xGA. It also allows us to see whether more xG or less xGA would lead to a greater shift in expected points.

Here is a quick run-down of the situation for Cambridge United right now, given 1.14 xG per game and 1.26 xGA per game. That comes out to 1.38 expected points per game, about 63-64 points over a season, which is likely to land a team around 11^th in an average League One season.

This output is telling us that, given our current xG and xGA, increasing our xG while keeping our xGA stable would improve us more than if we kept xG stable and decreased xGA. Thus, improving our attacking output might be the fastest way to try to move up the table. But again, this also hinges on not conceding a lot more goals than expected in games.

Thankfully, the team has already been increasing its xG pretty steadily since November. We can see that in the xG per game graph below. The dotted line represents the 4-match moving average xG.

Given that the January transfer window is now open, we can look for a few players who might be able to help either increase United’s xG or just convert the xG we’re generating. Obviously, the squad is thin up top from injuries and/or poor performers, so a striker should be the top priority. I don’t see an excuse for the club to not sign a new striker this month.

Striker Search

We need a striker who will score, simple as that. They don’t even necessarily need to create their own chances, as the team isn’t that bad anymore at creating chances. They just need to be able to finish their chances consistently that the team helps create for them. United hardly ever lead in games, so if Stevens (or another goalkeeper we bring in) is able to help us not concede and the club can bring in a healthy striker who can score, Monk can work on addressing that.

For this, I’ll use an app I built to scout players. Feel free to play around with it yourself! For a first sweep I’ll filter for strikers meeting a couple basic but important criteria:

1. At least 900 minutes played (10 full games)

Rationale: we need someone who has shown the ability to play games this season, but also, we need to ensure the sample includes enough players so this can’t be too high.

2. Top third for goal conversion rate (goals per shot on target)

Rationale: we don’t necessarily need to look for someone who has a ton of goals scored at the moment (as they might be expensive or play for too good of a team & not want to move into a relegation battle), but we need to find someone who can score the shots they do take. This helps us find decent goal scorers in poor teams.

3. Top third for aerial win %

Rationale: paired with a player who has a high conversion rate, a striker who is also good in the air can help as a midfield outlet and as a threat on corners, free kicks, and crosses… offensively and defensively.

I’m not including filters for things like xG, or number of passes, or shots, or carries. The reason for this is that we first and foremost need a goal-getter, and if they can help out in other areas it’s a bonus. Given the table position, I think the larger the sample the better.

I will also only look in a selection of leagues. I’m not planning on finding any possible loanee options from Championship or Premier League clubs because those wouldn’t really show up in my data filters, and might not have much career playtime to look at at all if they’re playing in Premier League 2, for instance. Plus, at the time of writing we’re currently maxed out on squad loanees. These are the leagues I filtered for:

League Two, National League, Scottish Championship, Irish Premier Division & First Division, Finnish 1^st & 2^nd tiers, Swedish & Norwegian 2^nd tiers

Through these searches, here is short list of strikers I think could be decent and maybe realistic options from these leagues.

All of these strikers meet our criteria, and might be players to look at for both the short term and for the next season or two. Below are a few data profiles for some of my favourites from this list.

Harry Smith, Swindon Town

Smith has proven himself as a capable goal scorer in a poor Swindon side this season. Good in the air, good at placing shots, and he already has some League One experience with Millwall in 16/17, Northampton in 20/21, and a very short Exeter spell in 22/23. He could be a decent option to look into.

Patrick Hoban, Derry City

Hoban was the joint-top scorer in the Irish Premier Division last season, able to finish in different ways and make his presence known in the box. He’s already 33 so might be more of a short-term solution, but I really do think he’d add some potency to United’s attacks.

Joacim Holtan, Kongsvinger

Holtan was on loan at fellow OBOS-Ligaen side Egersund for the 2^nd half of the season. He was a bit-part player for Kongsvinger and they sent him on loan, where he scored 6 and assisted 2 in 16 games. In 2023, the profile shown below, he scored almost a goal every other game as well and is very capable in the air. He gets into the box and is able to make the most of his fairly limited number of shots. This may be an “outside the box” player for a League One side, but I think United have shown that they often miss out on players who have multiple League One offers on the table. This could be a way to possibly be a player’s #1 option.

Justin Rennicks, Oulu

Rennicks is an American currently playing in Finland’s top division and has MLS experience with New England Revolution. Oulu have re-signed him ahead of the upcoming 2025 season after joining on a year-long deal initially in January of 2024, so they obviously like him – and for good reason. He’s a well-rounded striker with room to grow, has an eye for goal and has work rate to come back to help defensively. He gets in the box a lot but wasn’t the main shooting threat for Oulu, although I think United should take a look at him in a leading role.

Final Word

I’ve been writing this over a few days, and just as I finish up… the board release a statement saying that they’re backing Monk and will release some funds for key transfers this month. The hope is that the transfers the club splashes on are going to make an instant impact. They need to score goals, or at the very least help create them further back in possessions. While I think Stevens could be a good option in goal, perhaps they should also look for a goalkeeper who will save at least as many goals as expected – remember, our defence isn’t that leaky, it’s just that we’ve let in way more goals than expected!

I believe the club can sign at least 1 foreigner, so we’re not really limited to British players (I could be wrong here). Of course, the club needs to know about players to sign them. And then players will need to want to join a club who are looking like they could be in League Two in 6 months.

It’s difficult. The foundations and underlying data tell us Monk’s system should be allowing us to sit mid-table. But we’re 23^rd. In large part to our goalkeeper, in minor part to our lack of efficiency going forward… and with the caveat that we’ve virtually never had a healthy starting XI.

I’m looking forward to seeing what striker the club brings in. Because I don’t see how we can survive this season without a new striker, at the very least.