football stats- how they miss the point

Interesting piece from the Guardian on the use of stats to rate players performance. I agree with every word.
What do you think?

http://www.theguardian.com/football/when-saturday-comes-blog/2014/feb/03/statistics-football-analysis-miss-point-game

 

I'm not entirely sure what the article's point is. Obviously anyone relying entirely on stats for their opinion on a game or a player is a fool, but then nobody does that, do they? At the same time, anyone willing to entirely disregard statistical evidence in favour of their own, often misleading, memories is also kidding themselves. Disregarding stats entirely is just as daft as relying on them absolutely. There needs to be - and there is - a balance.

 

I agree with the article in that you cannot create an entirely accurate 'rating' system of players through stats alone and you absolutely certainly cannot create an accurate representation of how a game went by merely examining the stats. However stats used in isolation with the correct context are an excellent tool for the modern day fan. The problem occurs when you look past them being merely a guideline and start giving them precedence over what your eyes are seeing.

 

Ok Dan, in some sports I would agree. Cricket for example, there is no hiding from the stats.
Can't help wondering what the stats on Le Tiss would have told us that we couldn't have seen with our own eyes though? Would they have been used as a verification that he shouldn't have had more England caps? Stats are interesting to read but I am not sure that they add much. You think a player had a good game; the stats support your view, great. If they don't, do you say, "oh, he wasn't so good after all"?

 

The author is spot on. You do not need statistics to know that Cleverley is a shocking midfielder. On the reverse side Cleverley does need statistics to make out he is better than everyone thinks he is

 

Yes, exactly. "I thought Schneiderlin had a great game today, every ball he played found a target and he kept picking off balls through the middle." Look at the stats when you get home, "oh, he only completed 70% of his passes and only made one interception and one tackle. Guess he wasn't as good as I thought." Plenty of reasons for it, you may have only been paying complete attention to him when he was doing good things or may have confused him for someone else - happened with me when he played with Hammond all the time. Unless you're so stubborn that you're sure everything you pick up during one viewing of the match from an unideal position is spot on, there's nothing wrong with correcting yourself upon seeing evidence.

 

I take the point of the article but I'm not sure if he's trying to say stats are pointless (they aren't) or the use of stats can be improved (they can). It's not obvious to someone who has watched four minutes of MotD highlights how Zabaleta played high on the right wing for example, so some kind of image is good. You still need to use your brain for the context, but I'm not sure who is saying you don't. The thing about the "army of analysts" in the second last paragraph is garbage.

 

There are plenty of terrible opinions based on what someone saw or experienced (Guppy, Palmer, Seth Johnson - all called up to play for England). There are also plenty of people who are using stats who have no ****ing clue what they are doing.

 

I don't disagree with that, but a ten ft pass counts as much as a 40ft pass on the "pass completion stats" doesn't it? A full-back passing out to his midfield might have a great pass completion whereas the midfielder is trying to pass through the middle of the park which is much trickier might not. I really am an "evidence based" kind of bloke in my life, but I would only use the stats to slightly modify my view, not decide that I was mistaken.

 

Well, exactly, that's why total reliance on stats is ridiculous. If you hadn't watched the game, you might look and go "right, well Morgan completed 70% but Cork completed 90% so Jack was passing a lot better", when anyone who watched the game might have seen Morgan's 70% were pinged 40 yarders onto Adam's foot and through balls sliding Jay into the box, with the 30% incomplete being because of great defensive efforts, while Cork's 90% were five yard passes to guys directly behind him with the other 10% being playing opposition strikers through on goal. That's why complete reliance on stats could, in extreme situations, be objectively very, very wrong.

I agree with the article's viewpoint on completed dribbling percentages or whatever, though. I'm sure there are some games where Lallana has a poor dribble percentage because every time he gets the ball he's got three guys running at him from different angles, but a couple of times he still manages to sneak out from between them. Plus who knows what actually constitutes a dribble and all that - it's one of those stats I'm happy to disregard.

On the whole, it's not often I'll completely change my viewpoint based on statistics - the example I mentioned in my last post of the stats directly contradicting what I thought I'd seen is one example of when I will. Often, though, I'll happily be somewhat swayed by them - I'll come home thinking Lovren had a good but not great game, and then I'll see the stats and see he made 5 tackles, 4 interceptions and barely lost a header and be swayed to think he must have been even better than I gave him credit for. Especially because I regularly get him and Jose confused

 

Stats in any field are useless without the ability to analyse them. A player who plays safe passes all game will have a very high completion ratio, but a very low "key passes" or "chances created" score, which are arguably significantly more important. Perhaps the problem is not the use of stats themselves but that the stats which are given most prominence are the wrong ones.

 

But this in itself is a perfect example.

'Moneyball' was so interesting because it was looking at a sport already obsessed with stats. Problem was, people were looking at the WRONG stats. Hence why pass completion is not useful in isolation, the calculations of average pass length must be used alongside, then you can see the degree of difficulty alongside the rate of completion (for what it's worth, Morgans average pass length was about 20-25ft against Fulham, mostly shorter passes although his longer ones were all successful).

 

The problem is that there are so many stats around that can mislead as much as anything else. There is the one you always see in the CL when a player is subbed which shows the distance covered.

In the case of Morgan it is that he is all over the pitch. In the case of Makail Smith it is because he chases and harries non stop. In the case of Cleverley it's because he is the preverbial 'headless chicken'.

It would also make Jan Molby and Le Tissier look bad when the truth of the matter is that those 2 players did not have to move at all to do their jobs. Jan Molby used to just place himself on the centre spot and dictate the game from there whilst eating a pie and sipping on a Carlsberg. He was still a darned good player for it

 

Really pisses me off when the following happens:

You attend the game, describe to the board your thoughts on an individual players performance, tkaing into accoutn what he does both with and without the ball and somebody(suaully the same person who has not attened the game!) comes on here and says "3 ball recovries, 85% ppss completion rate - had a blinder, you're wrong."

Point is that stats do matter, they are underlying performance indicators. But the performance of a player is not always represented perfectly by stats. A player, for example, can have a 90% pass completion rate but not attempt any forward passes.

Conclusion: They are important but can be enitrely misleading - don't obsess over them!

 

You can look at the passing charts to see where every pass was made, where it went, and whether it was completed.

I would agree that stats are not that useful in saying which individual player had the best individual game-- for the same reason your eyes are not, either. The player could have had a tough assignment that day, or the pitch was bad in his area, or even just a bad bounce or two can change a performance. The sample size is too small. You could also argue that which player performed the best in a team game is meaningless no matter how you decide.

Over the course of an entire season, statistics do tend to more accurately capture how a player plays. And if you look at an entire season and ALL the players on the team, you will have a fairly complete picture of how the team played-- whether they were good or bad, the style they played, the strengths and weaknesses, etc. Much better than you could ever do relying on eyes and memory.

That was how Moneyball worked as well. None of those A's players were very good either statistically or scouting-wise. They were mostly one-dimensional guys, but they did little things well that typically aren't important-- like draw walks. There is no question you would rather have Barry Bonds on your team than Scott Hatteberg.

It wasn't that drawing walks was such a great skill compared to hitting 70 HR. It was that walks were undervalued in the market. They were worth a slight amount but people acted like they were worthless, because who remembers who drew a walk? And one player drawing a walk is actually worthless because someone has to be able to bring him home. But if you put in a lineup of 9 players who can all draw walks a little better than most then over 6,000 plate appearances it adds up.

 

I would suggest that Man City relied entirely upon stats when they purchased Edin Dzeko. No doubt someone can present some stats to demonstrate why my argument is wrong though!

 

The idea of "attending a game is the most valid way to have an opinion" has always interested me. I attend basically every single home game but very few away games, and I usually find I'm more comfortable in my opinion if I'm watching on a stream than if I actually go to the match. I don't think going to the game means in any way that your opinion should count for more.

 

Strongly disagree with this. If you attend a game you can see the whole pitch and not just focus on what the player does when the ball is at his feet which, in turn, allows you to see how theyr ead the game and speed of thought before and after they receive the ball.

 

You both have a valid point. I agree that being at the game, you see the bigger picture (if your seat is high enough in the stand) and when watching on television/stream that you don't always see that, but the flip to this is that at the game I often find that I get carried away with the match, only see the instance once in real time and can get wrapped up in the atmosphere not to have a clear thought on the game.

I agree that stats are dangerous and need to be taken as a small part of a performance rating. For example, I would hope that a winger or creative attacker would try and take on his defender a few times and even if he fails 80% of the time, it only needs one successful one to be done correctly to get that winning goal and hence job done.

 

My only concern with this part is the type of player. Eg. Holding midfielder who never takes a player on versus attacking midfielder whose job is to create something and go at people. The latter is more likely to have higher stats on giving the ball away and losing possession than the former over a season. The latter wins you games and the former stops the opposition winning games. How do you differ between the two in terms of performance based on stats?