In a regular column, Matt Rhein will look at the numbers beyond the results in the SPFL Premiership and, most of all, for Celtic. One of the very few (maybe only) analysts that track and share expected goals data publicly for Scottish football, in this first edition he looks at what the metric actually means and how it indicates an improvement in the shots taken by Celtic this season.
The feeling around Celtic Football Club has sure changed from this time last year. It is probably not fair to compare Ronny Deila and Brendan Rodgers’ reigns, as the type of players brought in and financial backing these two managers received are unquestionably very different. Regardless, there seems to be a confidence around the club that things are going in the right direction. What exactly has changed in the few months since Brendan Rodgers has taken over? Comparing the stats from this year and last, Celtic have arguably improved the quality of the shots taken.
The idea that a goal is more likely to go in based on the location and type of shot taken is the central idea behind the metric ‘expected goals’, also commonly referred to as ‘xG’. There has been some criticism of the idea of expected goals, but that seems to come from a misunderstanding of what exactly it entails. Very few would disagree that you are more likely to score a shot from 12 yards out taken with your dominant foot than you would on a shot 24 yards out taken on your weaker foot. If you agree with that, then congratulations, you believe in expected goals!
While the idea of analytics and expected goals is gaining more mainstream acceptance, there are still a lot of misconceptions about it. Even those who use the stat regularly would agree the name leaves something to be desired. Don’t we always expect goals?!
Along with the name, measuring expected goals based on shot locations could miss some things. For example, if a winger sends in a low cross across the penalty box that the striker is inches away from getting the required touch on that would steer the ball into an empty net, that would be an event on the pitch that you would think would indicate a good performance by a team. While most models would not record such an action, there are work being done on creating expected goal models that would include other such non-shot events. Nils McKay has worked on a possession based expected goals model that is very interesting, but sadly something like that is not possible for the SPFL as data publicly available for Scotland is very limited.
While the expected goals metric do have some limitations, it is becoming increasingly popular due to the fact that it has proven to be very useful indicator of a team’s performance beyond the actual result of a game. There have been numerous articles discussing how expected goals is a better predictor of future than actual match results (see here and here for further reading). Last year, the top six clubs in the SPFL had the six best expected goal differences (expected goals for minus expected goals against) in the league.
I have been measuring expected goals for the SPFL Premiership two seasons now (all data can be found on my blog: The Backpass Rule) and it can be used to confirm the notion that Celtic are taking better quality shots this season than they were last year. Celtic averaged 1.96 xG per match last season, while this year they are averaging 2.59 xG per match. Celtic is taking more shots this year on average, 17.94 to 16.81 shots per game, which can increase xG numbers. However, despite averaging more shots per game, Celtic still have a higher xG per shot this year compared to last with an xG per shot of 0.14 to 0.12.
Last season Celtic averaged 2.45 goals per game in the league. This is a very respectable number, but thus far under Brendan Rodgers Celtic has averaged 2.89 goals per game in the league. The reason behind this increase seems clear and thanks to the expected goals metric, quantifiable: Celtic are taking better quality shots than last year, leading to more goals.