Get Your Stats Right: January 2025

Looking at the numbers behind our new January signings

Well that’s January done and dusted. Eventful on and off the pitch, culminating in a deadline day that was not as eventful as anyone would have wanted. But over the course of the window, United did bring in five new players, so on the first GYSR of 2025 Ben G takes a look at their numbers, including a new metric that might sound complicated, but actually results in some pretty good news.

---

Goalkeeper Signing(s)

November’s GYSR: I discussed at some length how I think Reyes was a big problem for us. He wasn’t all that great of a shot-stopper, letting in a lot more than he should.

December’s GYSR: I touched on how Stevens playing (or “another goalkeeper we bring in”) should help us claw back some points because he saves about the same number of goals he’s expected to. I.e. he’s not under-performing like Reyes.

January 14, 2025: Marko Maroši signed.

January 18, 2025: 2-2 draw at Bolton. Maroši saves 1.88 goals above expected.

January 25, 2025: Maroši injured in the first half against Mansfield, before even facing a shot.

February 3, 2025: Nathan Bishop arrives on loan from Sunderland.

This is an absolute whirlwind of goalkeeper ups, downs, and sideways over the last 2 months. We go from Reyes, who in my opinion was the individual to shoulder the most responsibility for our poor concessions record, to Stevens who is an average shot-stopper but a bit better than Reyes, to signing Maroši who is an average to above-average shot-stopper, back to Stevens after Maroši gets injured after 1.5 games, to signing an emergency Nathan Bishop who appears to be a below-average shot-stopper (he was in his most recent starting gig, 21/22 Mansfield in League Two).

Personally, I’m thinking Stevens will be our #1 until Maroši is fit again, which to be brutally honest I’m ok with. Given the issues we had under Reyes it won’t be a total negative to have a better shot-stopper consistently playing, even if Stevens’ shot-stopping is nothing to write home about. He will have Bishop in training pushing him, and Bishop might be able to claim that #1 spot, which I’d be fine with too. I just don’t know enough about Bishop in 24/25 to pass a judgement if he or Stevens would be better until Maroši is back.

Regardless, here are the most recent statistical profiles for each of our now four goalkeepers this season. Reyes’s profile was fine on the ball but bad at stopping shots, Stevens is pretty much an average goalkeeper overall (which is not a slight on him, as we are rock bottom, so an average GK is great for us!), Maroši has declined a little over the past few seasons but last year was still an average to above-average shot-stopper, and then Bishop was similar to Stevens this year during the 21/22 League Two campaign.

And at the risk of both overloading you with stats profiles and rubbing salt in your “we have another injury?” wound, here are Maroši’s last 4 seasons.

We can see his decline from saving 0.12 more goals per 90 than expected in 20/21 and 21/22 to saving no more than expected in 22/23 and 23/24. Given this trend and his injury, it’s safe to assume he’d be around Stevens-level shot-stopping.

So, we have 3 average shot-stoppers, one with loads of League One experience but on the decline and freshly injured, one who has bided his time on our bench this season behind a below-average shot-stopping Championship loanee, and one who is an emergency Championship loanee who hasn’t been a senior starter since 21/22.

Have I made your day yet?!? 😊

Honestly, I truly do think that any of these 3 other goalkeepers will help us claw back a few points. And we have been doing that, with a 1-2-1 record worth 5 points in our last 4 games. Yes, 7 goals conceded in this span averages 1.75 per game which is effectively the same as the 1.79 per game we conceded before these 4 matches… but still. As I noted in November, conceding 2 cheap goals your goalkeeper should be saving lowers the mentality than conceding 2 goals that the goalkeeper did their best to save. And maybe we’re seeing that now. Plus, we scored first against Bolton, Mansfield, AND Rotherham! 3 games in a row where we scored first! Not conceding a cheap goal early on helps lift the rest of the team (or rather, instead of lifting them, it doesn’t bring them down).

Outfield Signings

We also signed 2 (count ‘em, two) strikers on loan from Premier League clubs, one of which is Dom Ballard who has actually been decent and rounded overall, although in true Cambridge United form he’s heavily under-performing his xG. At Blackpool before coming to the U’s, he recorded 4.7 xG and scored just 1 goal.

Here are all of Ballard’s shots in League One this season. He scored his goal on his 8^th shot of the season and hasn’t scored in his last 15. Hopefully he can improve, because something I really like about his profile is that he’s not afraid to shoot with his left or right. His shots come from about 50% left, 40% right, and 10% head. Having the ability to shoot with either foot adds an extra layer of unpredictability for the opposing goalkeeper, and allows him to maximise the xG of a chance since he won’t necessarily need to take an extra touch to get the ball onto his favoured foot.

Outside of goal-scoring, Ballard has also been one of the best attackers in the league so far at helping his team’s “underlying” performance. Using something called VAEP (valuing actions by estimating probabilities), I’ve created a model that scores each player’s actions in terms of the change in probability of scoring or conceding after an action. In short, this is the underlying “quality” of each action a player performs, from passes and shots to pass receipts and tackles. The brilliant data science academics at KU Leuven have a must-read interactive article explaining this better than I can.

My model also uses an algorithm to determine what type of role a player actually plays, which is usually better to use than team-sheet positions when ranking players against others. This is because you may have a fullback who pushes so high up and so consistently that they are, in actuality, a winger for 70% of the match. My model has Ballard as a winger, which makes sense since he drifts wide as a striker and he also played a lot of his Blackpool minutes as a winger.

The profile above shows Ballard’s main contribution is in receiving, shooting (i.e. xG, which we saw in his Wyscout profile earlier), and on-ball defensive actions. The receiving-shooting pairing is, to me, incredible to see because that shows he should be a dangerous outlet to send the ball to and get shots off. Even if he doesn’t score himself, maybe a rebound falls to a teammate. Ballard is average for the quality of his passes, dribbles, and loose ball recoveries, and below average in the air. I’m actually quite happy after seeing these numbers!

Below is a chart of all his actions that go into this model, colour-coded with the action quality score:

Overall, Ballard’s VAEP per 90 is in the 98^th percentile for all players in my model, with a VAEP/90 of 1.07, which is 0.44 above the average for wingers. This can be a little difficult to interpret, but essentially means he helped Blackpool +1.07 goal difference per game. However, each position varies and those playing closer to the goal tend to have much higher absolute numbers than defenders. So, a team’s total VAEP in a game can still be negative if they’re expected to lose. The key to focus on is Ballard has been one of the best wingers this season in terms of underlying action quality, so the lack of goals doesn’t worry me as much as it might for another player.

To compare, Josh Stokes is United’s “best” player by this metric, and ranks as one of the best attacking midfielders in the league. His 0.96 VAEP/90 is 0.36 above the average for his position, and ranks him 97^th percentile. So, we’ve basically added a player who can help us about as much as Stokes has…

… and what’s even better is that Stokes behind Ballard would theoretically elevate both of these players. Especially since Stokes ranks 97^th percentile for pass quality and Ballard is 95^th percentile for receiving quality. This could be a really dangerous pairing for us and help us get create more chances. And if Ballard can improve his poor finishing, even better. He should allow us to play more on the front foot, although the jury’s still out if he’ll help us finish the chances he gets or helps create.

Our other strikers and wingers (according to my position-role model, and with at least 540 minutes played) are shown below. Ballard has better action quality than any other players we have in similar positions to him, which makes me very excited about his loan.

Shane Lavery

VAEP/90: 0.87

Vs. Avg: 0.30

Dan Nlundulu

VAEP/90: 0.80

Vs. Avg: 0.22

Ryan Loft

VAEP/90: 0.69

Vs. Avg: 0.11

Elias Kachunga

VAEP/90: 0.64

Vs. Avg: 0.07

Brandon Njoku

VAEP/90: 0.48

Vs. Avg: -0.10

(My apologies for all the complex graphics & info… Trying to get at some underlying data rather than just looking at shots/goals!)

We also signed Esapa Osong on loan from Nottingham Forest, who played pretty well in Premier League 2 last season and, when comparing their profiles, it seems extremely similar to Ballard’s 24/25 (except Osong has less xG and more goals, better aerial win rate, and far more tendency to carry the ball, everything else is eerily similar).

And finally, we signed centre-midfielder Ben Stevenson who, much like Forest Green Rovers, was very good in 21/22 but then pretty iffy in 22/23. This is one of those cases where his “bad” data in a very bad 22/23 team isn’t a negative mark on him as a player. His team copped a -58 goal difference and averaged 0.58 points per game. He earned a move to Pompey that hasn’t worked out, but he hasn’t gotten a ton of game time, so we’ll see how he slots in at the Abbey.

Rounding Up

Overall, I’m not as disappointed in the January business as I was expecting to be. I wanted a goalkeeper who can stop shots and we got one (but he got hurt so we had to get another). I wanted a striker and we got 2, one who I think will be a solid addition and another one that adds depth and looks decent at Premier League 2 level. We also brought in a midfielder who has looked quite good in the past and perhaps a fresh start and consistent minutes will help him.

I’m a little unsure if either striker will be the answer we need to directly finish chances, but we aren’t under-performing our xG much at all. Ballard coming in to help us create more team xG should help a lot. We’re mainly under-performing our xGA by insane rates (14 more goals conceded than expected per Opta, by far the worst in the league) and I think much of that has been down to goalkeeping, so Strang and the club moving to address that, and backed up by data that’s a bit better than Reyes’, is uplifting.