Nothing. That’s the answer I’m sure many of you will give. The xG table is a fanciful, made-up table that has no meaningful information at all.
If you think that, I feel you are wrong. However, you’re also wrong if you think it is the be-all and end-all. It’s like possession, shots, and all the other numbers that make up a game of football – stories are contained within, but they’re not the whole narrative.
However, I couldn’t help but be drawn into a debate on Twitter the other day around Mark Kennedy and Michael Skubala. The discussion was around Kennedy’s football. There’s a misconception, spread by perhaps one or two people, that he was told to play a certain brand of football. Obviously, that is rubbish because if he was, I’m almost certain that football would have been more entertaining than it was. Mark switched up his style a couple of times, trying to be attacking, going defensive, then playing three at the back and looking to play counter-attacking football. It worked in many games, and we finished 11th, a good place given how tough things were for a while around Christmas.
That said, we weren’t at the races when it came to xG. I’m not going to explain what xG is again, but I believe there are stories you can take from it. My belief in xG as a performance guide started in 2017/18 when Plymouth Argyle finished seventh. Their xG data was appalling; they should have been 16th based on xG but finished seventh. I distinctly remember George Elek saying how they’d be in trouble if they didn’t improve, and 12 months later, they were relegated. That sold me. It’s true that xG, as a one-off, can only suggest where changes need to be made, but if the warnings are not dealt with, you can find yourself in trouble.
Cue the first table – last season’s xG. Now, before you look at this, the table I present to you is the outcome of the league if, instead of goals scored, games were decided by the xG result. So, for instance, we won 2-0 at Wycombe last season and therefore got three points. The xG data suggests that in terms of the quality of chances, we should have lost 1-0 (xG was 1.79 for them, 0.75 for us). This model takes all of the results across the season and reimagines the league table on xG outcomes. It’s better than the out-and-out xG table, especially for us, as we were the fourth-lowest for chances created last season – only Shrewsbury, Morecambe, and Forest Green were worse. That doesn’t take into account the defensive aspect, though, whereas this table does.
Aside from the fact that we are a lot lower than 11th, what’s interesting here is the placing of some of the other teams. Wycombe, Charlton, and Fleetwood are all worse than their xG, and guess what – all three have struggled this season. Oxford, who finished 19th last season, should have been eighth based on xG alone. Tell me, how are they doing this season? The pattern is there; for some teams, Shrewsbury and Burton, the problem has persisted, and so has the outcome.
The point is, whilst we finished 11th and drew positives from that, there was always a nagging doubt in my mind (which I vocalised on the podcast) that we needed to see a shift in approach; we needed to create more chances. One user on Twitter suggested that the first six games of the season were evidence we were doing that, but actually, not a huge amount had changed. We lost the xG battle against Bolton, as we did the game, and whilst we beat Wycombe 3-0, the xG suggested 1-0 was about right. Our 2-2 draw at Northampton saw us get battered in terms of xG (0.98 to 2.95), whilst Shrewsbury should have narrowly beaten us (1.24 to 0.97), and the 3-0 Blackpool hammering was closer to being 2-1 (2.02 v 1.38). Finally, the Bristol Rovers game that finished 1-1 was spot on in terms of xG. So, from the first six matches, from which we took 11 points, the xG suggests we only deserved seven. Seven from six? Not so impressive.
We stopped the debate at six matches because that’s when the injuries kicked in, and those defending Mark Kennedy say the data is skewed. That said, if we do fast forward, we only won two games on xG (Peterborough and Portsmouth) and should have lost against Carlisle and drawn with Cheltenham.
Still, those numbers are skewed by the lack of forward; I’d tend to agree with that argument, but if we fast forward, what can we glean from Michael Skubala’s last six matches? We’ve taken six points from them, a relatively low return, although we have played Derby, Peterborough, and Blackpool in that run. However, the xG suggests we’ve been much better than the points tally suggests. We should have beaten Northampton 1-0 (1.05 v 0.66), Wycombe 2-0 (2.04 v 0.92), Peterborough 1-0 (1.52 v 0.68) and drawn with Burton and Derby 1-1. The awful showing at Blackpool, a 2-0 defeat, was spot on. So, from a run of six games where we have six points, the data suggests we should have taken 11.
Other Skubala games where we should have got three points and didn’t include Barnsley (2.22 v 1.54), Wigan (1.04 v 0.2), and Derby (1.09 v 0.95) away. How does that make the xG table look as of Tuesday morning (without the midweek games)? In fairness, this doesn’t have Wigan and Reading’s point deductions in, but it looks a little something like this.
Yep, on the balance of results, we could/should (depending on your view of xG) be ninth. I’m not saying we should be, but I am saying that there’s been a real improvement in our chances created and the defensive work. We’re conceding less xG and making better chances going forward. We should be four points better off, and when you imagine that under MK, we were in a deficit, it shows the huge strides the club is making.
Now, if I were a fan of Shrewsbury and Burton, I’d be worried because there’s been no upturn in their xG output over the last season and a half of this. I’d be worried if we’d also been around the bottom four, because eventually, your luck will run out. Having dug into the numbers, I firmly believe that had we not pulled the trigger in October, we would be hanging around that bottom four, in xG at the very least.