Most of the people who have commented on my recent statistics blog post, both here and on Twitter, seemed to agree with it, though there were a few dissenters. Some have pointed out that Arsène Wenger himself is a big fan of stats, so how dare I criticise them, and by implication him. Others say that although it is clearly more difficult to pinpoint useful stats in football because of the fluid nature of the game, that’s no reason to stop trying. Some thought that my logic meant that I was saying all stats were worthless, or even that I believe stats should be able to help us predict the future. Maybe they weren’t reading it properly. The only thing you can definitely predict about the future is that it’s unpredictable. Anyway, in the words of Jules Winnfield, allow me to retort.
To repeat the main point of my argument, statistics are useful if they’re in context. I’m not saying that football managers or anyone else should stop using stats or looking for new things to measure to try and gain an advantage. Obviously I put a slightly provocative headline in order to be slightly provocative. Really I meant ‘An Undetermined Percentage Of Statistics Are Stupid And Pointless When Taken On Their Own Without The Surrounding Context’, but that didn’t fit the page so easily. Nor would it attract the same attention.
Tim Stillman (aka @LittleDutchVA) made some good points to me on twitter about Arsène’s use of stats in coaching, looking at such things as how long players take to release the ball, how many interceptions defenders make (attempted and succeeded), and how these things and many more can be used to build up a picture if used scientifically. It should be obvious that ratios can be useful in coaching: successful passes to overall passes; successful shots to overall shots; same with tackles, saves, pretty much anything. And I’m sure Arsène does try and use all these scientifically, unlike most fans, who just quote random isolated stats that may well be curious or mildly diverting, but don’t really tell you anything about a player or a match.
However, in football even putting a load of stats together as scientifically as possible is still not the whole story, because so many actions are either unmeasurable or never repeated in the same way. Suppose a player moves two yards to his right because he anticipates what an opponent is going to do with the ball; the opponent then changes his mind and plays a different pass to his original intention; the outcome of the move is different. Now we will never know how different in real life. Maybe this one action has made the difference to whether a goal was scored, whether one point or three was taken, whether a team is relegated or wins the Champions League. Yet it is most unlikely to feature in any list of stats because the action itself can barely be measured, never mind the consequences. If we had an infinitely large computer then an infinitely large number of scenarios could be played out and we might have an idea of the effect of this one action. But similar actions and decisions are taken all over the pitch for the whole 90 minutes, so just to analyse one match in this way would take more computing power than the human race currently has at its disposal. Football is chaos. It’s random. It’s like the weather. Weather forecasting has got much better over recent years because computer models get bigger and bigger and different scenarios can be played out millions of times to see the most likely outcome. But until we can track every molecule in the atmosphere we can’t be sure of our weather predictions except short term and large scale. Footballers are like random molecules moving around a space, and just as hard to measure properly.
There are actually very few sports as random as football, where anyone can go anywhere on the pitch at any time, and the ball can also go anywhere as long as it stays within the lines. Even rugby has rules about the direction of the ball, and far stricter offside laws. As far as I know rugby is not particularly heavily analysed for stats. There are obvious ones to do with points scored, successful kicks at goal, scrums and lineouts won, but not numbers of yards gained as in American football. Is there an equivalent in football (‘soccer’) to American football’s yards run or thrown by a player? I don’t think so, because American football is highly regimented – less so than baseball and cricket, but far, far more than football. In American football you run and throw in one direction; to go the other way almost always damages your team’s chances. In (real) football there is often an advantage to going one step back before you try and take one forward. Maybe even three steps back.
Think about this: suppose Arsenal have 60% of the possession in a game (so the opposition have 40%, obviously); Arsenal have a pass completion rate of 76% (opposition 64%); shots on target 12 (opposition 3); shots off target 8 (opposition 5); corners 9 (opposition 0). None of that tells us who won, but worse than that, it doesn’t even tell us who played better. Probably Arsenal did, but not necessarily. Their on target shots may all have been feeble, while their off target efforts went out for throws. Their passes and possession might have been down to playing the ball around at the back and timewasting. I don’t think there is another sport where these kind of misleading statistics could even occur.
Or compare two players, let’s call them A and B, or perhaps Arteta and Bendtner. Let’s say in one game Arteta had a pass completion rate of 86%, made 78 passes, one interception, two shots on target, two tackles, got tackled twice, played two through balls and didn’t get booked. Bendtner had a pass completion rate of 68%, made 49 passes, no interceptions, one shot on target, one tackle, got tackled three times, played no through balls and got booked. Who played better? I’ll tell you: we don’t know. We don’t know because even with all those figures we’re not getting enough context. For one thing, we have no idea what anyone else on the pitch was doing when all the above stats were being racked up, and how they were influencing the events. We’d either need about 5,000 other statistics to be sure of what was going on, or we could just watch the damn match. Because maybe watching the damn match tells you more than statistics. Just a suggestion.