Expected goals, witchcraft and bean counting

Slowly, football is starting to catch up with other sports – many across the Atlantic – in its use of statistics to analyse the game.

One of the key metrics to surface in the past few years is expected goals (xG), which has grown to the point it is now in the Match of the Day stats reel at the end of each game.

But there has been a fair amount of push-back against xG at times, as people react to something that supposedly taints their image of the ‘beautiful game’.

So let’s start with how xG works. If you’re reading this, you probably already know, but to summarise quickly in the words of Opta: “expected goals (xG) measures the quality of a shot based on several variables.”

The accuracy of different models produced by different companies/people will depend on what variables are included in the model. Opta advise that they use assist type, whether it’s with the head or foot, and angle/distance from goal as part of their model.

It is not a stat made of fairy dust, part of a wizard’s staff and other mystical items. It is looking at historical data on shots, and seeing how likely a chance is to be scored based on that.

And given that football is about about scoring goals which tend to come from shots/chances, it is relevant information.

Some of the push-back against the stat seems to be surrounded by claims it is sucking out the emotional element of sport, and stopping us from working things out and understanding what has happened from watching the game.

But this is absolutely not the case. It’s a stat and, much like any other, in some situations it can be used to shine a valuable insight on certain things.

Take Colchester in League Two this season. For roughly the first third of the season, they were well in a play-off battle. But their underlying xG numbers were very poor, and eventually their poor chance creation compared to opponents has caught up to them. They now lie in 21st.

Or Mason Greenwood, perhaps? He has struggled for goals this season after a stellar 2019/20. Look past the provocative tabloid headlines that came out in Autumn, and this dip in goalscoring could probably have been predicted given he scored a startling 10 goals from 3.39 xG (according to Understat) last season.

It indicates last season was more of a hot streak thanks to his excellent finishing ability, and the striker is yet to consistently find himself in good goalscoring positions.

Of course, xG isn’t perfect and, like all other stats, there are some situations where it doesn’t quite work out. Brighton are the big example at the moment, having vastly underscored their xG over the past couple of season.

And back down in League Two, you have Cambridge consistently outscoring their xG to send themselves top of the league. They don’t particularly look like relenting having maintained this all season.

But when the stats don’t match up to the results on a consistent basis like this, it’s not a reason to discard or dismiss that metric. It’s just an indicator to dig a little deeper to try and understand what’s happening, be that with a different set of stats or with a good old fashioned eye test.

Stats like xG are a great way to try and understand a game better and look for underlying trends. But they don’t have to be engaged with if you don’t want to.

If you love xG and possibly do a lot of work with it in media or for a club, that’s fine.

If you like watching the game then checking out the xG and comparing it with what you saw, that’s fine too.

And if you have no interest in that kind of thing and simply just like to watch and not get involved with xG, then you know what? That’s fine too!

Nobody is arguing that stats should be looked at entirely instead of the games themselves. But viewed alongside, they can provide interesting and sometimes telling insights.