I recently started watching basketball, and one of the things that struck me was how important it was to get a shot away, as opposed to having the ball stolen, and the importance of converting your attacks. In a high-scoring game like basketball, it seemed that it wasn’t so much a case of having to out-score your opponents, but kind of out-converting them (these are the same in essence, but feel different to me, I guess, semantically).
Could this be applied to football? Basketball is obviously a higher-scoring, higher-percentage game than football (higher-percentage in that each shot in basketball has a higher percentage chance of being converted than in football), and the nature of attacks is quite different. The idea that you should try to convert all of your attacks into shots is a far, far cry from analytics at the moment, where differing quality of chances is clear in the data.
However, there was something about the idea of converting possessions to attacks, and attacks into shots that stuck, and seemed feasible within the footballing context.
I ‘got’ numbers of possessions for a team by adding up all of the things that would end a ‘possession’: missed passes, shots, opposition defensive actions. There would be things like running the ball out of play too, but in the grand scheme of things I think these are relatively infrequent so wouldn’t affect things too much by their absence.
For ‘attacks’, I adapted Sam Gregory’s definition that he used for his Opta Pro proposal and article (http://www.optasportspro.com/about/optapro-blog/posts/2015/guest-blog-is-a-%E2%80%98one-dimensional%E2%80%99-attack-such-a-bad-thing/) and took it as anything that ended in a shot or an opposition defensive in the last 20 yards or so of the pitch*
NB: Both of these mean that one single spell of ‘possession’, as we might recognise it visually, could be split into multiple different ‘possessions’ or ‘attacks’, eg if a cross is headed out by a defender but falls to an attacker who shoots, these would be counted as separate attacks.
To take an example to focus on, Arsenal ‘convert’ possession into attacks 37% of the time, which is the second highest in the league, suggesting that when they get it, they’re dangerous. They then convert these attacks into shots 28% of the time, which, again, implies a dangerous streak. In defence though, their opponents convert possessions into attacks 35% of the time, which is pretty high compared to where they are in the table, it’s higher (so worse) than the league average of 32.5%. However, their opponents then only convert their attacks into shots 16.6% of the time, the lowest rate in the league. This seems to suggest that while they may be fairly open when they first lose the ball, they’re good at shutting things down where it counts.
At Chelsea, they’re converting possessions to attacks 34% of the time, which is slightly above average, and attacks to shots 22% of the time, their figure nearly bang on the league average. Given the high conversion rates that City and Arsenal have, I think these are related in some way to quality/form, so perhaps the proximity to the average here indicates Chelsea’s troubles this season. Defensively, their opponents are only converting possessions to attacks 30% of the time, better than the league average (it’s possible a low opponent possession to attack conversion is an indicator of pressing too, though I haven’t investigated this). However, of these attacks, over 25% of them are converted into shots, which is one of the worst in the league.
I’m not entirely sure how useful these kind of statistics can be. They may be able to indicate some things but they’re largely linked to shot numbers (shots per possession% – what percentage of possessions result in shots – particularly). I think it might be best described as extending shooting conversion backwards, so that whereas before the chain was “shot -> shot on target -> goal” it’s now something approaching “possession -> attack -> shot -> shot on target -> goal”.
I’ve also got the ‘passes per possession’ numbers, which I guess are a rough indicator of the speed at which teams play (although I think there are other methods to measure speed of attacks/play out there). Perhaps unsurprisingly, Louis van Gaal’s sleepy army top the list. I’m not sure what opposition passes per possession indicate yet.
Of course, to get rid of problems of schedule it would be better to see these for an entire season as opposed to just part of one (these are figures for the first eleven games), but I thought it would be more interesting to have more immediate relevance as a starting point.
As an added bonus, if you’ve got to the end, here are the numbers for how many attacks per game a team features on average, on both sides of the pitch. No prizes for guessing who’s top.
NB: As the previous NB mentioned, an ‘attack’ as we as a viewer might see it can be split into multiple ‘attacks’ here, so Leicester matches don’t go from end to end every 45 seconds or so.
*’20 yards or so’ because my data is taken from the StatsZone website relative to the website pictures, rather than the official Opta data itself, and as so the distances are approximate (which is partly why I didn’t wish to use the Sam’s definition of actions inside the penalty area as well as that would have been an approximation too).