Measuring Football

[rcoobc]

You can prove anything with statistics. 60% of people know that.

But is measuring football useful? If we are simply using numbers to confirm what we already believe, we are blatantly and openly experiencing confirmation bias. Especially when we are also happy to dismiss statistics that don't agree with our viewpoint: e.g. that a player passes backwards (dismissing a high pass success rate), or is afraid to take long shots (dismissing a high chance conversion rate), etc.

Let's look at a few examples

2002-03 - Strikers

If we look into the annuals of the Premier League, 2002-03 is an interesting case.

Three players chasing the Premier League golden boot, and it was even closer that that going into the last two games. Ruud scored 4 times in United's last two matches, Henry once (despite Arsenal scoring 10, and three players getting hat-tricks), and Beattie failed to find the net at all.

So it's clear, Ruud was the best striker that season? Well hold on one second, let's look at assists. Beattie got 2, Ruud got 4, whilst Henry got a staggering 20 assists!

But what about penalties. It doesn't take much skill to score one of those, and Henry didn't take them? Actually Henry took all but one of Arsenals converted penalties that season (I don't know how many were missed). In total, Henry scored 3 penalties, Beattie 4 and Ruud 7!

So what do we really learn from this?

RVN scored 33% of United's goals that season (25/74). But only contributed 4 assists.
Henry scored 28% of Arsenal's goals that season (24/85). But contributed a staggering 20 assists!
Beattie scored a crazy 53% of Southamptons goals (23/43). But only contributed 2 assists.

I don't think any of these numbers really teach us very much about the players.

2016-17 - Playmakers

But maybe looking at goals and assists to assign a value to a player is very last century.

Let's look at "key passes per game" from the last season

No major surprises there. The lack of a United player on the list possibly tells you more than the list itself.

But does those numbers alone tell us anything interesting about the player? How many times did they try the killer pass and fail? What are the players passing accuracy? Why are Hazards assists so low? Can we build a useful metric if we take into account those sets of data too? Why does Ozil have half the number of assists compared with De Bryne despite have a higher Key passes per game? Does an unconverted killer pass indicate something else too?

The point of all that is to lead to a simple question;

What stats do you think are a fair metric to judge a player in any position? Are there any stats you could say before-hand are a clear indication of a players quality? Could you build a model (the scientific kind) of attributes that clearly show which are the best and worst players (before knowing who each player is).

Or, are you a "there is only one number that matters in football" type person.

 

[Cascarino]

Interesting OP.
Personally, I think most fans who use statistics use them erroneously at best. You get a lot of people who'll google goals scored/assists made and use it to make up their mind on a player, rather than actually watching them. You'll get someone arguing that player A is better than player B because he has more tackles per game, completely ignoring things that can't be quantified.

Statistics can be useful when done in conjunction with watching the player, giving a more holistic view of his playing style and abilities. Especially with the increasingly thoroughness of statistics.

 

[Stack]

The trouble with many stats is they dont show the entire picture and in some cases get used out of context of a situation.
A really good one is something you mention with respect to players playing backwards a lot. Lots of people will criticise a player for constantly playing backwards or sideways. What doesnt get looked at is why. There are some common and simple passing routines teams use to move the ball and opposition players around. Those dont get quantified by stats. Sometimes a player will move into space, receive the ball (usually on the back foot) and then pass it sideways or backwards. What doesnt get picked up by the stats is that may have been deliberate to open up a passing lane so the next pass can go forward. Some players are awful at simply passing back to retain possession but the point is stats dont show or differentiate with respect to this and this is just one example.

 

[Akshay]

What stats do you think are a fair metric to judge a player in any position? Are there any stats you could say before hand that are a clear indication of a players quality? Could you build a model (the scientific kind) of attributes that clearly show which are the best and worst players (before knowing who each player is).

I think to get any useful information at all you have to use a bundle of measures, not just one. To use your key passes example, there are plenty of players who have a higher key passes / 90 mins but have lower pass completion or vice versa. So one isn't necessarily a better passer than the other, just more adventurous or playing further up the pitch. Key passes isn't a great stat anyway as it's only measuring passes leading to a shot, but says nothing about the quality of that shooting opportunity. You'd need loads of stats (and better ones than we get on squawka) to assess a player's performance in a single game.

Then you have the problem of judging that performance in light of the strength of the opposition, which is a whole other can of worms. Then the problem of assessing a player's quality from the performances. Bad performances doesn't necessarily mean poor quality, could be a tactical system letting a player down, being played in wrong position/role, insufficient support from teammates etc. We're very far away from any scientific model.

Right now I prefer to trust my eyes for 80%, and check very particular claims from statistics. For example, in discussing Herrera we can determine if he's an erratic passer (has high pass completion so no), mainly passes sideways (more forward passes so no), doesn't contribute towards build up of goals (has a good assist record for a central midfielder so no), or is erratic with forward aerial balls (yes, he has a poor completion % in this category).

 

[Keeps It tidy]

You should take as many things into account as possible

 

[JPRouve]

The trouble with many stats is they dont show the entire picture and in some cases get used out of context of a situation.
A really good one is something you mention with respect to players playing backwards a lot. Lots of people will criticise a player for constantly playing backwards or sideways. What doesnt get looked at is why. There are some common and simple passing routines teams use to move the ball and opposition players around. Those dont get quantified by stats. Sometimes a player will move into space, receive the ball (usually on the back foot) and then pass it sideways or backwards. What doesnt get picked up by the stats is that may have been deliberate to open up a passing lane so the next pass can go forward. Some players are awful at simply passing back to retain possession but the point is stats dont show or differentiate with respect to this and this is just one example.

Exactly often stats are used in isolation without taking into account teammates and managers even though they are crucial to how a player will act on the field. Iniesta is an example of that, if you have never seen him play, know the context around him and just base your evaluation on stats, you will most likely think that he is unimpactful.

 

[Thunderhead]

aye, David Silva is often criticised for not getting enough assist, but he's generally playing the ball into a channel for someone to then get the 'assist'

for me assist's should only be used for FFL, not the actual worth of a player

 

[ValenciasDrilledCrosses]

aye, David Silva is often criticised for not getting enough assist, but he's generally playing the ball into a channel for someone to then get the 'assist'

for me assist's should only be used for FFL, not the actual worth of a player

And assists are only counted as assists if the person on the end of it has the skill to finish the ball. So it could be a great pass made but the player won't get the assist because the player on the end of it completely fluffed the chance. The assist stat bothers me probably most of all out of all stats people cite in arguments or discussions

 

[Minimalist]

The point of all that is to lead to a simple question;

What stats do you think are a fair metric to judge a player in any position? Are there any stats you could say before-hand are a clear indication of a players quality? Could you build a model (the scientific kind) of attributes that clearly show which are the best and worst players (before knowing who each player is).

Or, are you a "there is only one number that matters in football" type person.

I think to answer to your final question: the truth will (as usual) end up in the middle somewhere. I generally think that statistics are great at providing clues and trends that we can take advantage of. Statistics are only as good as they are applied as most people know. I've been dumbfounded learning more and more about stats during my PhD research, at how badly they can be applied to argue something that the data just doesn't show.

I think any metric you can think of must be used in conjunction with another measurement (or some context). Example, 20 assists in the PL is usually valued more than 20 assists in the French League. I mean they do this with the golden boot/shoe don't they?

I think a general measurement (despite how obvious it might be) is simply the success of a team and a players role within it. If Eden Hazard is involved in 80% of Chelsea's goals and Chelsea win the league, then aye, he's a good player. That probably sounds remarkably simplistic but that's the real low hanging fruit we have. Or slightly less obvious, if a centre back of a team that has survived relegation (having been nailed on for it by the bookies) is making more interceptions and completed tackles than 75% of his peers in the league, then he's probably worth a look.

As for creating a model, that's bound to be in development already. There would be too much money on offer to any company that could provide proper predictive modelling for their players/team. I do think football has far too many moving pieces to make this an easy feat though.

 

[Mibabalou]

I agree with OP somewhat, yes these statistics don't show the whole picture but no club is going to wiki to look up a players total goals. The analysis that clubs will perform use more in depth stats ariel dules won, passing stats left vs right feet, distance covered, sprints, stuff you can't discern just by watching games. Don't for one second think scouts and managers just sit on excel all day looking for gems stats are normally a supplement to watching players not all they do.


Arguing with people on twitter yeah sometimes stats is all they got but let's be serious no one wins when you argue with someone on the internet

 

[Peyroteo]

If you know how things like 'key passes' or 'chances created' are measured then it's tough to take them seriously. Even the number of assists can be very misleading.

Stats are very useful and it's why top clubs invest a lot into it, but you must actually watch the games and understand the context of those numbers to be able to take any conclusion from them.

 

[edcunited1878]

If you know how things like 'key passes' or 'chances created' are measured then it's tough to take them seriously. Even the number of assists can be very misleading.

Stats are very useful and it's why top clubs invest a lot into it, but you must actually watch the games and understand the context of those numbers to be able to take any conclusion from them.

Bang on to your second paragraph...

Lads, numbers without context is just numbers.

 

[11101]

I read an article recently that was talking about how clubs are now looking at 'Expected Goals' as the main stat that helps them identify transfer targets. It was saying the most closely related stat you can find on squawka or whoscored is shots per game from inside the box.

 

[rcoobc]

I think a general measurement (despite how obvious it might be) is simply the success of a team and a players role within it. If Eden Hazard is involved in 80% of Chelsea's goals and Chelsea win the league, then aye, he's a good player. That probably sounds remarkably simplistic but that's the real low hanging fruit we have. Or slightly less obvious, if a centre back of a team that has survived relegation (having been nailed on for it by the bookies) is making more interceptions and completed tackles than 75% of his peers in the league, then he's probably worth a look.

That's largely where I am as well.

Statistics like key passes are *interesting* but I don't they should really be used to win arguments. Even goals scored and assists, I think, are on dodgy grounds.

Take Leighton Baines, in 2010-11 he played every minute in the Premier League, scoring 5 league goals and coming 5th in the leagues assists charts. By 2013 he's creating more chances than anyone else in top five leagues (according to wikipedia), and he's a defender. So why has Real Madrid not picked him up for £80m?

I think if you asked most football fans if they think Baines would honestly improve their team during that time, the response would probably be "yes, but not by much". Baines normally played with Pienaar in front of him. (1) (2). Towards the end of that period, the effectiveness of Baines began to wain, and may actually have become a liability for Everton. Still he is now the defender with most assists in Premier League history.

In passes, key passes, assists, goals scored, etc, you aren't going to find a better full back than Baines in the Premier League. But how does he affect the rest of the team? Does he make the team better? Would he make Chelsea or City or Real Madrid better?

I'm doubtful

 

[rcoobc]

I read an article recently that was talking about how clubs are now looking at 'Expected Goals' as the main stat that helps them identify transfer targets. It was saying the most closely related stat you can find on squawka or whoscored is shots per game from inside the box.

That certainly makes some sense.

 

[AXVnee7]

Fans tend to use statistics to defend a player or back up their argument, and disregard them when the stats go against them.

 

[HTG]

Almost all stats, as long as they are collected or made carefully and without mistakes, are useful when used the right way. But context is always necessary and there is obviously no perfect stat.
But the ones I personally like the most are packing and goal impact. Especially the goal impact. GI is as objective and simple as it gets and when you look at the players who are best at it, the stat seems to work pretty well.

Oh, I hate it when people start arguing about how much ground a player covers. The stat can be useful, but is horribly overrated.