For many years Total Shots Ratio has plodded along as a good indicator of team shooting performance, not overall team performance, but shooting performance.
It’s a good enough indicator that its found its way into generic match reports for professional soccer teams and has good visibility on Opta – a well recognized soccer statistics company now owned by Perform Group.
But with all that publicity and ‘useability’ that doesn’t make it ‘right’!
Why do I say that?
Within a game of football there are always two teams playing against each other – so team performance statistics should not only take into account what the attacking team is doing – they should also take into account what the opponent is doing to the attacking team.
So what do I mean about modernizing TSR. Most define TSR has simply the volume of shots one team takes versus the volume of shots another team takes. That’s okay but the end state is excluded – the result – a goal scored.
So my new vision of TSR centers around the end state as well as the volume – in other words the equation for Attacking TSR (ATSR) now becomes Goals Scored/Shots Taken and then Defending TSR (DTSR) becomes the percentage of your opponent’s Goals Scored/Shots Taken.
Finally, in looking at how well Composite Possession with Purpose correlates to Points Earned in the League Table I would create Composite TSR (CTSR).
Before getting to the numbers – some history first:
I built Possession with Purpose using this philosophy and if you’ve been following my efforts for the last two years you know that my correlations to points earned in the league table are extremely high… To date:
- MLS 2014 = .86
- Bundesliga = .92
- English Premier League =.92
- La Liga =.91
- UEFA Champions League =.87
So let’s peel back the regular way TSR correlates to Points earned in last year’s MLS – when viewing the old way (Total Shots only as a percentage for both teams) the Correlation Coefficient “r” for the entire league was .32.
My new way of calculating CTSR with the End State of Goals scored has a correlation coefficient “r” of .75
Far higher… now for some data.
Here’s the correlation of the my new TSR Family of Indices shows with respect to Points Earned in the League Table – the same analyses used with respect to CPWP above:
- MLS 2014 ATSR .74) DTSR (-.54) CTSR (.75)
- Bundesliga ATSR (.53) DTSR (-.41) CTSR (.68)
- EPL ATSR (.86) DTSR (-.35) CTSR (.76)
- La Liga ATSR (.88) DTSR (-.77) CTSR (.92)
- UEFA ATSR (.64) DTSR (-.40) CTSR (.65)
Like CPWP the correlations vary – in four of five competitions the CTSR has a better correlation to points earned in the league table – while in one case (the EPL) ATSR has the best correlation.
So how do the numbers stack up for some individual teams when evaluating ATSR, DTSR, CTSR, and CPWP compared to those teams points earned throughout the season?
In other words what do the correlations look like (game to game) through the course of a season for sample teams within each of those Leagues?
In almost every sample TSR (now ATSR) has a lower, overall correlation to a teams’ points earned in the League Table than CTSR (Borussia Dortmund and Barcelona being the exception) – this pattern follows the same pattern seen with CPWP almost always having a higher correlation than APWP and Goal Differential almost always having a higher correlation than Goals Scored.
I’ve also taken the liberty of highlighting which Composite Index has the best correlation to points earned between all four categories – in every instance either CTSR or CPWP is higher than TSR. But, as can be seen, sometimes CTSR is higher than CPWP…
What this proves is that there simply isn’t one Index that is far better or far worse than the other – it shows that different teams show different styles that yield better relationships to points earned in different ways —> meaning there is not only room for improvement in current TSR statistics but room for the inclusion of PWP principles within the Industry standard.
I would offer – however – that even when you create CTSR the backbone of that data can’t offer up supporting analyses on how a team attacks or defends. It’s still only relevant to the volume of shots taken and goals scored.
And while the volume of shots on goal and goals scored appears to be a constant across most competitive leagues (average greater than 5 and 2 respectively for teams winning on a regular basis) the average of shots taken for winning teams is not as constant… (Expected Wins 4) —> why I favor PWP over TSR – nothing personal – just my view…
I’m not sure I did a good job of comparing what I view as the old way to calculate TSR (the way that ignores the End State of Scoring a goal) and how an update to it can help tell a better story that actually correlates better to the complexities of soccer.
COPYRIGHT, All Rights Reserved. PWP – Trademark