The variety of practical applications for Expected Goals in football

Those of you well versed in the football analytics movement probably already know that Arsenal are the major club considered to have the most progressive analytics department, having purchased their own data company, StatDNA, a couple of years ago, with Arsene Wenger even namedropping Expected Goals a couple of times in the past season. There have also been growing rumours that analytics is playing more and more of a role in their player recruitment with recent signings like Gabriel and Mohamed Elneny being mentioned as examples of that practice.

Expected Goals served as something of a deflection for Wenger when first choice midfield duo Santi Cazorla and Francis Coquelin were injured almost simultaneously. He had used Aaron Ramsey on the right flank during the early season, partly to squeeze him into the lineup, but also because Ramsey lacks both the technical consistency and the defensive nous to play in midfield without a partner who can both defend and build play from deep. With both Cazorla and Coquelin out for months, Wenger’s hand was forced, but instead of communicating that, he decided to focus on what could be expected from Ramsey’s stint in the middle:

“Aaron is a player who creates a high number of chances when he plays through the middle and his expected goals when he plays through the middle is quite high, one of the highest in the Premier League”

Whether Wenger’s conclusion is correct or not (that Ramsey’s high ExpG numbers from the middle means he should play from the middle) is questionable. Football is a dynamic game of so many moving pieces that focusing too heavily on one piece can mean that the rest of the puzzle falls apart, as it ended up doing in the case of Arsenal this season. Wenger, obviously, knows this, and as was explained above, moving Ramsey inwards was predicated more by a lack of other options than by anything else. It is, however, interesting that Wenger brought it up in a press conference because it suggests that it is something that affects his decision making. This is something that the season’s Expected Goals numbers verify: Arsenal led the league in Expected Goals and Expected Goal Difference throughout the season, even if they couldn’t translate that into tangible success.

The reason why I find it interesting is because the biggest hurdles for concepts like Expected Goals is for it to be taken seriously in the places where it really matters. If you’re a manager and you want to implement Expected Goals in a way that affects the way your team plays, you first need to communicate it to your players in a way that makes sense for them, either through actually explaining the concept or by training in such a way that the message gets through. The same problem can be found in the media, where conservative pundits have reacted to the new metric with the kind of scorn usually reserved for terrorists and murderers – which is pretty funny when you consider that it’s just a number. Having someone like Wenger use the term, however tentatively, scores it believability points that adds some serious gravitas – it’s no longer exclusively something for spreadsheet wielding nerds.

It’s also a concept that tries to explain some of the most basic, conservative thoughts you’ll hear a football pundit utter: “they’ll be gutted to have lost that, they deserved to win it”. The only problem seems to be the inherent objectivity, being able to say with certainty that one chance was better than the other, or that either of the teams created the objectively better chances. The concept is inherently accepted, through years of playing and watching and debating football, but the metric is new and scary and therefore it must be shunned. Yet it shouldn’t be shunned, because it has a practical application for just about everyone involved in the game, whether it be players or pundits, scouts or coaches. If communicated properly, it can lead to more intelligent football discourse, more efficient recruitment, smarter strategising and better on-pitch actions. Just listen to Mesut Özil describing a chance he created for Aaron Ramsey this past season:

I don’t think Wenger’s fascination with Expected Goals has anything to do with the video, nor do I believe that the pass that Özil describes was influenced by Wenger’s preachings – like the tweet says, the guy loves to assist – but it still serves as quite a perfect illustration of why Expected Goals matter. As Özil himself puts it:

“I could have shot myself but it would have been a 50/50 chance, because I would have had to shoot with my right. Aaron was free, I saw him, played the ball to him and he scored. The chance of us scoring was then 100 per cent.”

His numbers are crude but the concept is solid: if I shoot I have a 50% chance of scoring, but if I play the pass which I expect to make 80% of the time, I can increase the chance of scoring to 100% (or thereabouts). Or, we can break it down further: the 50% comes from location on the pitch, which foot the shot would be taken with, the shot being created by a through ball, and the 100% comes from location, there being a through ball in the buildup, which foot the shot would be taken with and the location of the goalkeeper. It’s super simple, and that’s what makes it illustrative.

Now, this is only an example, and one where the calculations are really simple, but the concept is equally applicable in different, more complex, scenarios – as long as you’re capable of efficiently doing the kind of crude number crunching presented above, which is where Expected Goals can help you. If you have a clear sight of goal from outside the box on your weaker foot and your alternatives are to shoot, to pass or to dribble, your instinct might tell you to take a shot, but if you know that the ExpG value of that shot would be, say, 0.03, it might make you reassess your options. Or if you know the respective values of a cross compared to a cutback, or an in-swinging corner compared to an out-swinging corner, or a counter-attack compared to an attack based on a slow buildup. It allows you to make better, more sophisticated decisions in situations where you might not have the time to do the calculations by yourself. Or, if you’re a manager, it can allow you to focus your resources on the things that are truly valuable, rather than wild goose chases.

Pundits should embrace Expected Goals for what they are: a deeper shot statistic that offers the kind of context that mere shot numbers can’t. It can aid the discussion about the game just as much as it can aid what happens within it, it can offer a clear and precise answer when you’re scratching your head, asking yourself how the team with only 30% possession managed to win that game. And that’s all Expected Goals, as a concept, is: a way to describe things in order to make them more understandable and a way to answer the question, “should I shoot or should I pass”.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s