Marco Silva: The xGenius

Imagine it’s Friday night and you’ve just been to see a play in the West End. You had a great time and it got a standing ovation from the audience. On your way home, you check twitter and see that a complex mathematical algorithm has assessed the play, and decided it was actually rubbish. The probability of a standing ovation (xO) was only 0.38, so the reception it got was more than it deserved…

Anyone would dismiss this as bonkers because theatre is soobviously subjective, which brings us to the point of this strange metaphor. What is football? Is it an art to be enjoyed and disagreed on, or a science to be number crunched and solved? In the “old days” the beautiful game was definitely viewed as more of an art than a science, but the rise of statistical analysis and data modelling has challenged this perception. With sophisticated enough modelling, couldn’t team performance be quantified and boiled down to a simple stat?

The answer in my mind is no. Or at least, if such a thing is theoretically possible, we are a very long way from achieving it. And yet lots of commentary around football is intent onusing xG as if it is some sort of shorthand for who ‘deserved to win’. A group called ‘The xG philosophy’ – which has hundreds of thousands of followers on social media – publishes something it calls The Justice Table, depicting what the league table would have been if points were decided on xG alone. This approach is nonsense. Having a higher xG does not necessarily mean you were the better team, or that you deserved to win. It is footballs most wildly misinterpreted metric, and I’ve had enough.

Given that our first game of the season was yet another example of Fulham winning the match but ‘losing’ on xG,let’s unpack what it is, what it tells you, and what its limitations are. xG (or Expected Goals to use its full name), is a data driven metric which predicts the probability of a shot going in, based on the quality of the chance. That is all. It says nothing about how you found yourself with the chance, just the chance itself.

To explain how useless this is at showing who deserves what, consider this scenario: Your midfielder smashes a volley intothe top corner from 30 yards – probably gets an xG of 0.05 for it. Five minutes later, the other team does the same thing, except their 30-yard volley bounces down off the crossbar and gets tapped over the line. Their xG will be way up at 0.90 because the final chance was a tap in. But does that mean they ‘deserve’ more? Clearly not.

There are loads of examples like this. A jammy penalty will earn you about 0.80xG, but an amazing team goal will get you diddly squat unless the final touch is on a plate. Some xG models even double count the same goal – so if you have a shot saved and your team mate follows it up, the team getsmore credit than if you’d just scored from the initial chance. Come on people. Come on.

For very specific uses, xG can be helpful. Say you’re a scout and want to compare how good two strikers are at finishing their chances – xG could be helpful for you. The problem isn’t the existence of xG in itself but the way it is interpreted as some sort of holistic ‘deservedness’ measure. If we are going to shoehorn loads of data into football, we should at least try and understand what it all means accurately. Match of the Day and Sky Sports do very little explanation, they just stick it on the screen. I would argue bombarding people with numbers serves no purpose unless you explain what you’re showing them.

Is there a place for data in football? Sure. We’ve used data for decades – tracking possession, counting shots, corners, throw ins, you name it. But people understand what these stats mean, and so no one goes around making a ‘Justice Table’ for what the premier league would look like if it was sorted by number of corners. If anyone tries to make you believe xG is a magic measure of who deserved to win a game of football, they are just showing a lack of understanding of our favourite sport, and how far away we are from making it more science than art.

As for the Everton game, we did genuinely ride our luck a bit… broken clock is right twice a day and all that. But hey, wehave super Bernd Leno in goal for a reason, and I’ll happily take the smash n’ grab.

So up the Fulham, and here is to another season of upsetting the xG brigade by outperforming what we ‘deserve’.