The craft behind football’s advanced statistics

Image: Pexels

Expected goals – what’s that? The mainstreaming of advanced analytics and metrics in football.

This summer marked a significant moment in the increasing popularity of football analytics, with the announcement that expected goals (xG) would appear on Match of The Day for the 2017/18 season. This gives xG a four million viewer audience per week – a number far greater than the niche segment of twitter where I first discovered it. Even though it is rarely spoken about on the programme, the exposure is a substantial change.

So, you may still be wondering what xG is. xG is a stat composed to quantify how good a shooting chance really is, so that you can tell if your star striker really should have scored that chance in the last minute to win. Statistic company ‘Opta’ are one of the leading proponents of xG. By analysing over 300 000 shots they assessed the likelihood of a shot going in and applied a numerical value ranging from 0-1, with a higher number representing a better chance.

The value also acts as a percentage value, so if a shot has a value of 0.1 it will be scored 10 per cent or 1 in 10. The value considers several factors to provide an accurate numerical reflection of the chance if taken by an average player. Opta’s official diagnosis for coming up with a quantitative objective metric includes: passage of play (open play, free-kick etc.); assist type (type of pass provided); off the dribble; rebound; header; distance from goal; angle of the shot; one on one/big chance; and finally competition adjustments for a subset of competitions.

As we can see, the model considers a varying degree of factors in order to provide an accurate reflection of the chance at hand. So, while it’s all good discussing numerical values of shots, how useful is this to the wider aspect of football?

Some critics of the model have argued that it’s a generally useless stat. They have suggested that applying a value to shots is useless; why should you care about goals that are expected to be scored when we have a numerical value for the goals
that have actually been scored?

Well, xG is an extremely useful metric that you can measure over the long term. It can tell you whether your team’s recent form is legitimate and thus is it also sustainable, as most teams or players that exceed or under-achieve their expected goals often regress to a mean of a sustainable level.

Two great recent examples of this are Juventus’ slow start to the 15/16 Series A season and Leicester’s surprising rise from relegation battlers to title winners. But first the case of The Old Lady’s slow start in the quest to retain the Scudetto. Juventus started their first ten games losing four, drawing and winning three apiece. Opta had their xG at 19 while they had only scored 11, creating a -8 differential; similarly their xG conceded was 5 yet they had conceded 9 through 10 games creating a -4 differential. Thus, we can see that they were simply not finishing their chances while simultaneously conceding too many. Eventually Juventus would regress to their expected level and even further, winning 25 of their next 26 games as they went to retain the Scudetto for the fourth consecutive year.

Conversely, the case of Leicester City was one that shocked the football world, marked 5000-1 to lift the Premier League title. For the 15/16 Premier League season, Opta have Leicester conceding 10.7 goals less than their xG conceded were for the year. This is significant when you consider that 14 of their 23 wins were decided by one goal. While not the whole explanation, this somewhat helps to explain their rise and subsequent decline, which saw them flirt with relegation and sack their title-winning manager, just six months after leading them to one of the greatest underdog stories in football history.

Finally, the use of xG in the wider context of football should be used only in conjunction with what you see on the pitch. Stats provide great arguments to back up what you have already seen. It is an incredibly useful tool to spot and recognise patterns in a team’s performance, whether they’re sustainable or not. However, they should not be used as the only measure to judge your team’s performance, as this can be very reductive.

Leave a comment



Please note our disclaimer relating to comments submitted. Please do not post pretending to be another person. Nouse is not responsible for user-submitted content.