Maybe I’m wrong and it’s always been this way, but it seems to me that there’s a trend for football journalists to make more use of statistics in their reporting of matches. And not just simple reporting of match statistics, but a much deeper insight into how the fundamental data from matches can add insight that either reinforces or contrasts with the viewable action that actually took place on the pitch.
As a great example of this, I’ll disentangle below bits of the match report of last season’s Everton versus Chelsea Premier League game by Jonathan Liew of the Guardian. For context, Everton had been poor all season, without any appreciable improvement in performance since replacing Rafael Benitez with Frank Lampard at the start of 2022. At the time of their home game against Chelsea on 1 May, they had just fallen into the relegation zone, 5 points behind Burnley, albeit with two games in hand. So they were a big club with big problems. The Abromovich situation also complicated Chelsea’s run-in, but Jonathan’s article is all about Everton.
I’ll just highlight the most relevant parts from a statistical angle. Again for context, this game ended up being quite bitty. Everton didn’t create much, but they won 1-0 with a goal by Richarlison just after half time. Basic statistics show that Chelsea dominated possession and created more chances than Everton, but watching the game it never really felt like Chelsea were going to score.
So, here’s a breakdown of the relevant parts of Jonathan’s review:
Richarlison knows that his chances of winning the ball, on his own, against a Champions League‑winning backline, are almost non-existent.
This frames a footballer’s decision-making as a process involving risk assessment. What are the chances? (In this case, of winning the ball and getting a chance to shoot.) What’s the reward? (In this case, a goal, and a potential victory.) How motivated is the player? Does the potential value of the reward compensate for the low chance of success? The fact that Chelsea have a strong defence – evidence: they won the Champions League the previous season – reduced Richarlison’s chances of even winning the ball, never mind scoring, to virtually zero.
So go the margins. Might Everton still have won this game had Richarlison not conjured a goal out of his own irrepressible will? Maybe, even if they had 22% possession and barely a third of the shots.
Some basic match information here: very low possession for Everton and far fewer shots than Chelsea. Which tells a story. But the bigger story is that Everton won. And as also expressed, football is a game of fine margins. This isn’t explicitly a statistical comment, but since football is a low-scoring game compared to most team sports, a small gain – whether in performance or luck – can easily make the difference between winning and losing.
Might they (Everton) still have retained their Premier League status had they not won this game? Maybe, even if a draw would have left them four points behind a rampant Burnley side.
This provides statistics about the game in its context of the Premier League table. The game wasn’t exactly ‘must-win’ for Everton, but even a draw would have left them with a much bigger tasking of chasing Burnley with a four-point deficit (with a game in hand).
But like pretty much every challenge Everton have faced this season, you would have to bet against them.
So, both in this game and throughout the season, you would ‘have to bet against them’. Though not to be taken too literally, I interpret this as a statement about Everton under-performing relative to their ‘big team’ status. As it says elsewhere in the article, Everton are supposedly ‘too good to go down’, a perception that’s likely to influence market prices, whereas the reality is that they were generally playing for most of the season like a team who could very well go down. This mismatch between expectations and reality may plausibly lead to a bias in the market, leading to value on bets against Everton.
In a way this has been Everton’s predicament all season: a tyranny of low percentages, of attacks that are doomed to fizzle out and a defence that will inevitably wilt sooner or later.
‘A tyranny of low percentages’ speaks for itself, both in football and statistical terms.
If you can’t keep the ball, and you can’t keep the ball out, and your main goalscoring centre-forward has been injured for much of the season, then you are relying on a lot coincidences to go your way.
In a nutshell: everything in football (and life generally) is signal and noise. And if the signal is low for any reason, you need to get lucky with the noise to compensate. Moreover, a football match comprises many, many chance events and a weaker team is likely to need a lot of them falling in their favour to overcome a deficit in skill.
And 45 minutes into this tense and occasionally tetchy game, Everton’s luck looked like it was running out.
First half summary: it didn’t look like Everton would get the luck they needed to overcome their deficit in quality.
The odds were against Richarlison.
This is a reference to Richarlison’s goal, which essentially came out of nothing when he put Azpilcueta under mild pressure on the edge of his box. Chelsea and Azpilcueta don’t make many defensive errors like that, so it was unlikely that the press from Richarlison would be successful…
But then, the odds have been against him before.
… while this refers to the start of Richarlison’s career, which is discussed earlier in the same article.
It was his 92nd-minute goal that rescued a point at home to Leicester two weeks ago. His equaliser that rescued a hopeless cause against Arsenal in December.
Some more basic data, this time to highlight the importance of Richarlison’s contribution to Everton’s survival chances across the whole season.
The statisticians will tell you that Everton got lucky here: outplayed in terms of possession and outshot in terms of chances, ransacking a win courtesy of a lucky deflection and an inspired goalkeeping performance.
This is actually stretching things a bit. As explained above, the statistics do point to Chelsea dominating the game – 78% possession, 17 -9 superiority in goal attempts. But shots on goal only favour Chelsea by 5-4. And though XG – which gives a measure of expected goals based on chances created – varies from source to source depending on method of calculation, at least one reliable source of XG made the game close to a draw on these terms (Everton 1.12 to Chelsea’s 1.21). So, a strict interpretation of the statistics does suggest a Chelsea win was more likely, but any football statistician looking at those data won’t be at all surprised that they coincide with an Everton 1-0 victory.
The counter-argument is that the last five games of a relegation battle is no time to be worrying about percentages.
And this final comment is really a statement about motivation effects, especially at the relegation end of a table. Any analysis of data that ignored the fact that this was a much more important game for Everton than for Chelsea would be likely to under-estimate the chance of an Everton win.
In summary, Jonathan Liew’s match report is littered with statistical references, each of which adds insight and context to the literal description of the match action. Bear in mind though that Statistics is never a precise science – it’s always about finding the most reasonable interpretation for events that were shaped, at least partly, by chance. So, be critical when reading reports of this type. For example, as I mention above, the statement “the statisticians will tell you that Everton got lucky here” is not untrue, but the statisticians are also likely to say that the result is not that surprising given the other match data.
Finally, while Jonathan is perhaps the very best football journalist at making use of Statistics in match reports, he’s definitely not the only one. If you come across similar match reports where the statistical element is either particularly interesting or central to the reporting, drop me a line and I’ll add a link to it here.
