Predicting football is at the same time virtually impossible and rather simple. For example, we can be fairly confident in predicting the teams that will be challenging for the title next season and which clubs are likely to be involved in the relegation dogfight. However, is there a measure that might give us a bit more objective predictive power to determine what might happen over the course of a season?
The excellent football analytics blog by James Grayson has looked at a measure called Total Shots Ratio (TSR). This is a fairly simple measure that looks at the ratio of the total shots in a match that each side has. In other words, the formula for it would be:
Total Shots Ratio = Total Shots For / (Total Shots For + Total Shots Against)
Applying this to the 2012/13 Premier League season that has recently ended, we can calculate the TSR for each club. To see whether there is any correlation between this value and the club’s final position in the table, we can plot TSR against points for each club.
Clearly, we can see that there is a reasonable correlation between the two. Indeed, the R2 value for the line of best fit in this chart is 0.582.
There are a couple of interesting anomalies in the data that are worth looking briefly at. Manchester United acquired a significantly higher number of points than the best-fit line suggested that they should. Their TSR of just 0.534 suggests that they should have collected 57 points, as opposed to the 89 that they did achieve. Similarly, QPR’s TSR of 0.459 should have returned them 45 points and comfortably kept them in the top division.
The clear flaw in this measure is the lack of any indication of the quality of the shot. A team could have 25 of the 30 shots in a match, but if all of their shots are from 30 yards while their opponent’s 5 shots are all from inside the six yard box, the TSR would not be a particularly great indicator for that match.
So, we could hypothesise that Manchester United may have not created as many chances as we might have expected relative to their opponents, but the quality of their shots were superior. Indeed, we find that Manchester United’s 37.9% of shots on target was the highest in the Premier League, which combined with the quality of their strikers might explain their TSR. Similarly, QPR’s 29.2% shots on target was the lowest in the Premier League – in other words, they may have taken plenty of shots, but not many of them hit the target.
It is worth looking at this over a longer period. With just 20 data points for the 2012/13 season, we cannot be confident that these results are necessarily correct. Having collected TSR data for the past eight seasons, back until the 2005/06 season, this could be plotted this against the total points.
We can clearly see a solid positive correlation between TSR and points, as shown by the R2 value. In the graph, the red data points indicate the team that won the title that season, the yellow points are those teams that qualified for the Champions League, while those in black are those teams that were relegated.
We can make a few observations from this. No team that has won the title has found themselves below the line of best fit for this data. While no champion has been quite as extreme as Manchester United this season, they all find themselves with a better than average TSR with respect to the number of points that they obtained. Similarly, albeit with a couple of exceptions, all the relegated clubs have been below the line.
This seems to suggest that while having a good TSR is certainly important, simply having a large number of shots without taking into account the quality of the shot is not enough. Teams can also improve their TSR in two ways – either by increasing the number of shots that they have or by restricting their opponents to a lower number of shots.
In the future, it will be interesting to look to incorporate a measure of quality of shot into this and see if we can use that to try and explain some more of the variation that we see in the data. However, we can see that TSR does provide a reasonable good predictor of a team’s performance over the course of a whole season.