The story so far:
I saw a tweet from Ted Knutson (@mixedknuts) saying Leicester were lucky in 2015-16 – specifically he believed they were lucky to concede fewer goals than his ‘expected goals against’ stats suggested they should.
I wrote a blog post saying I didn’t see any evidence of luck in that stat, and explaining why I thought Leicester’s tactics, teamwork and spirit were the reasons for their success. Even 11 ‘Galacticos’ can only get so far; it’s a team sport, after all.
I alerted Ted Knutson to my blog post, but had no response. However, after mentioning it again on twitter I got a couple of responses from ‘Mark’ (@ETNAR_uk) and ‘Joel’ (@messiseconds). I’d like to thank them for making the effort to reply, and this shouldn’t be seen as any kind of attack on them, it’s just my continued effort to get to the bottom of whether everything is being taken into consideration when people throw around the claim that Leicester were lucky. The league table was unusual and some of Leicester’s stats might well be unusual, but that’s not in itself enough to claim it was all luck.
It does seem Spurs fans are the most likely to shout about how lucky Leicester were, like a caller to 606, who got savaged by Savage – have a listen here. You won’t learn much about football or stats, but it’s funny anyway.
I’ve no idea if either Mark or Joel are Spurs fans, but I’m sure it’s not relevant to their comments.
As someone amongst stats Twitter who’s maybe a little less hardline, I think parts of [your first blog post on this] are largely sensible, BUT Leicester have def been lucky, to an extent. You mention so many variables in football and Leicester have had almost all of these gone their way. Some of your criticisms of xG (eg the set-pieces breakdown) are good, and things stats people also point out when they have the character space. Tonally, the dismissiveness is likely to piss off stats folk who get a lot of trolls in their mentions (which I don’t have enough followers for, which may be why I’m more likely to give more time). The amount Leicester over performed their xG is a level which is beyond what you’d expect to just be variance, though. And (I’m not around clubs so can’t say for sure) I highly doubt ANY coach ‘relies’ on stats, at most it’s a supplementary thing.
In response to that I’d firstly say thanks for acknowledging that some of what I say is largely sensible. Then I’d say: What have Leicester had ‘go their way’ that they didn’t make happen themselves? As I explained last time, they’ve used effective tactics, worked as a team and made themselves considerably more than the sum of their parts. That’s what smaller clubs do if they want to compete. Usually they manage it for a few games a season, and every ‘big’ club usually gets beaten at least once a season by a struggling team. The fact that Leicester have done it consistently over the season with very few poor performances doesn’t necessarily mean they were lucky. Perhaps it just means they worked harder to maintain a level of performance that is unusual for players on their wages.
Mark also said: ‘The amount Leicester over-performed their xG is a level which is beyond what you’d expect to just be variance’ – okay, if you say so, but just because something doesn’t fit a model, that doesn’t make it luck. Perhaps the model doesn’t cater for all circumstances. This is the real world. Things happen that are outside theoretical models – economists get caught by this all the time. Fund managers are notoriously useless at it. But if it is luck, please show me numbers and examples of that.
As for coaches relying on stats: well certainly not 100 per cent, but there are lots of claims these days that players were signed on the basis of stats, so there seems to be growing potential to be caught out by it.
There’s a tonne [of evidence for Leicester being lucky].
1) Your view of football seems deterministic, and we have a lot of evidence to suggest that’s not the case
2) A tonne of teams have outperformed xGD [expected goal difference] in fluke seasons, but they almost never (BMG) do. This is like 10 seasons of 10+ leagues. The narrative always over-fits. Is it possible that specific plays aren’t being picked up by xG? Yes but, again, that’s a) slightly unlikely b) very very unlikely to continue.
‘Determinism’, is the philosophy that all events have causes that are external to the human will. So if my view of football was deterministic, wouldn’t that mean that I was the one who thought it was all down to luck? That Leicester wouldn’t be able to influence events? Sounds like the exact opposite of what I’m saying. In my opinion their will played a large part.
Point 2: I’m scratching my head on this a bit. I did ask Joel if he could explain further, but he says he’s too busy due to exams until June (not sure if the exams are his or someone else’s, but it doesn’t really matter).
I’ve tried to interpret, and if he wants to come back at a later date I’ll be happy to update if I’ve got something wrong.
Firstly I don’t know what BMG is, apart from Borussia Monchengladbach, and there may be a mistyped or missing word or two in that sentence, which often happens on twitter of course, but isn’t helping me right now. However, the rest of it:
‘This [either Leicester winning the title, or just their xG stats – not sure] is like 10 seasons of 10+ leagues.’ Meaning it’s a one in 100 chance? In which case, I’d have to say: so what? One in 100 events sometimes happen. If Usain Bolt runs 100 races and in one of those he breaks his own world record, is that luck, or is that because he tried hard and eventually succeeded? Or if I throw darts at a board without even looking, one in 100 of those hitting the board might end up in the 25 or bulls-eye. Is that luck? No it’s just an event that sometimes happens when I throw darts at a board. On the other hand, if someone bets that my 85th un-aimed throw out of the 100 will be the one that lands in 25 or bull and it does, that’s lucky. The luck is in predicting when the one in 100 outcome will arise in an essentially random series, not in the fact there is one.
‘Is it possible that specific plays aren’t being picked up by xG?’ I don’t think I was in any way claiming that specific plays weren’t being picked up by xG, I was saying that individual examples don’t include enough detail to all be lumped together. For example, a shot from the exact same spot in the D of the penalty area is different – and therefore has a different chance of success – depending on (but not limited to) these things at the moment the shot was taken:
- which direction the shooter was facing
- what speed the shooter was moving
- which foot he used and which he preferred to use
- the level of skill of the shooter
- height of the ball
- speed and direction of the ball
- how much in control of the ball the shooter was
- the state of the pitch
- the wind speed and direction
- whether it was wet, damp, dry, sunny, cloudy, in shadow, cold, hot, etc
- where all the defenders and goalkeeper were at the time of the shot
- which direction, if any, defenders and keeper were moving in
- tiredness of both body and mind of every player
- how long the defence had to be ready for the shot
- where all the shooter’s attacking teammates were, what they were doing and how much attention the defenders and goalkeeper were paying to them
- the level of determination of every player between the ball and the goal to get in the way of the shot
You only need to change one of those to affect the chances of the shot resulting in a goal, and some of those things can’t be subjectively measured. So I don’t see how an xG model is ever going to include them – and even if it did, on some days a factor might increase the chance of a goal, while on others the same factor might decrease it, because of its interaction with another factor. So you can lump all those shots in together and average them out, but every now and again something will come along and appear to break the model, because the model isn’t detailed enough for all the possible nuances. That’s not luck, that’s an imperfect model.
Conclusion: no one has yet given me any evidence that shows Leicester were lucky. If you have some, please share it. I am genuinely curious about this, I’m just not convinced Leicester have been particularly lucky. I tend to think that over 38 games they will be as lucky or unlucky as any other team. In a way you can categorise everything in football as luck – no one got in the way of a shot, it hit a post and went in rather than out, because the wind was 10mph not 15mph, and the player who shot was only on the pitch because someone else got injured – so it goes on. In a way it’s ALL lucky, but in another way if you practise hard and work hard, the end result might go your way, so it’s not luck. Anyway, Leicester’s deviations from ‘expected’ stats are, in my opinion, due to their actions and the actions of their opponents and I don’t regard Leicester as being luckier than everyone else.
But I am open to having my views changed if there is a good reason to change them, so if you disagree, feel free to tell my why.