Judging players solely by stats

[stw2022]

It's very possible that a player will have played very well yet not have anything to show for it in the key stats metrics (at least the headline ones). Equally a player can easily have a poor game but pops up with a moment in the game that gets him something that reflects well on the stats. Over the last three or four years there seems to be an intolerance towards any suggestion that how a player has performed may have no correlation with the a small number of headline stats.

You get an assist = you've played well

You haven't had any key passes = you've played badly.

To the point where 95% of discussion around a player is now 'LOL you think he played well? Lowest key passes in the team" or LMAFO - yeah he played so badly, that's why he got an assist" - and that's the end of the discussion.

I've watched players I thought had stinkers who got an assist or a goal and players who haven't registered on any key stat metic who performed well.

Yet this is rarely ever accepted now. Everyone seems to watch a game and form an opinion on how well someone played based on the stats that come out of it. As if they can form no opinion watching the game.

 

[Pogue Mahone]

It's very possible that a player will have played very well yet not have anything to show for it in the key stats metrics (at least the headline ones). Equally a player can easily have a poor game but pops up with a moment in the game that gets him something that reflects well on the stats. Over the last three or four years there seems to be an intolerance towards any suggestion that how a player has performed may have no correlation with the a small number of headline stats.

You get an assist = you've played well

You haven't had any key passes = you've played badly.

To the point where 95% of discussion around a player is now 'LOL you think he played well? Lowest key passes in the team" or LMAFO - yeah he played so badly, that's why he got an assist" - and that's the end of the discussion.

I've watched players I thought had stinkers who got an assist or a goal and players who haven't registered on any key stat metic who performed well.

Yet this is rarely ever accepted now. Everyone seems to watch a game and form an opinion on how well someone played based on the stats that come out of it. As if they can form no opinion watching the game.

Definitely true that stats can mislead in an individual game (although not if you dig into all the available stats, stuff like possession lost, possession regained, number of pressurew blah blah) but they’re obviously much more relevant/useful when you combine data from lots of games.

 

[Scandi Red]

Stats can for sure be a conversation killer. The issue is that most people are scientifically illiterate (or whatever the proper term is). You can for the most part use stats to tell the story you want to tell. I think goalkeeper stats are particularly easy to tweak.

That does of course not mean that stats are useless. Generally speaking, if a player has lots of goals or assists then it becomes hard to argue against. But it should never be as simple as "Player A had 20 goals and Player B only had 15 goals, therefore the former had the better season". That is just one aspect that doesn't necessarily tell the whole story.

 

[Rado_N]

Definitely true that stats can mislead in an individual game (although not if you dig into all the available stats, stuff like possession lost, possession regained, number of pressurew blah blah) but they’re obviously much more relevant/useful when you combine data from lots of games.

Except xG, that can get right in the fecking bin.

 

[Skills]

Football fans are terrible judges of what's happening in front of them.

Numbers paint a much better picture even if they're incomplete. The examples you've used are poor though - a player not getting a goal or an assist doesn't mean they're having a bad game. Football is inherently a very low scoring game - so goals and assists are relatively rare for most players.

However, key chances, passes, progressive passes, receptions, pass completion all paint a fairly descriptive picture about what's happened over 90 minutes - especially if you compare them to other players on the pitch and those in similar positions.

 

[padzilla]

You can literally have an assist by passing it two yards and the person who receives it does all the work to score. Alternatively you could dribble past six players and set up an open goal only for someone to miss it. I'm not convinced assists aren't only a thing because of fantasy football.

That said if someone's racked up dozens of assists then it paints more of a picture.

 

[Skills]

You can literally have an assist by passing it two yards and the person who receives it does all the work to score. Alternatively you could dribble past six players and set up an open goal only for someone to miss it. I'm not convinced assists aren't only a thing because of fantasy football.

That said if someone's racked up dozens of assists then it paints more of a picture.

That's exactly the same for the goalscorer in each of those scenarios.

 

[Pogue Mahone]

Except xG, that can get right in the fecking bin.

I spent 10 minutes this morning wrapping my head around PSxG-GA. 10 minutes I will never get back.

 

[Scandi Red]

That's exactly the same for the goalscorer in each of those scenarios.

Honestly, when you start digging most stats are useless in isolation.

A key pass may still dependant on a smart run or movement around the box. Passing accuracy is largely dependant on playstyle, tactics and how much and how smart your team mates move. Saves depend on the quality of the shots and how good the defenders are. Tackles and interceptions can be a direct result of formation and how high the defensive line is. Etc etc.

 

[tomaldinho1]

If we had advanced enough stats it would be fine but the issue is we're miles away from that.

I will say xG and similar improves every season however as more and more data inevitably make sit more accurate but it is basic.

 

[Ish]

I spent 10 minutes this morning wrapping my head around PSxG-GA. 10 minutes I will never get back.

tl;dr version for the rest of us, please Pogue

 

[Superunknown]

I tend to find stats get used more when teams or individuals aren't doing particularly well. It matters less when things are going well, because you know the team is functioning. If you're doing well, then everything must be working well. It becomes more analytical when you're trying to identify why examples teams are underperforming or why the goals aren't going in, or why there are so many draws or losses as opposed to wins. That's when you tend to find that the stats are then used as a stick to beat others with. SkySports and all of these other broadcasters, Twitterfolk, etc, they hold on to these stats like they're the most important thing. They're useful, at times, when used appropriately and when used in the context of the game and what's happening on the pitch. The most important thing is to watch the games and see for yourself what's happening.

Stats can be useful within games. E.g. for the last 10 mins, this team has had 70% of the ball. This team hasn't had a shot on target for 60 minutes. It can be useful information if you put it in the right context. The most important aspect of the game continues to be goals scored, goals conceded, games won, games lost, and points obtained. The stats can quickly become a distraction.

 

[HTG]

If a player impacts the game in a positive way, this can always be measured. We might not have found all useful measurements yet, but that’s another story. In theory any positive impact must be measurable.

 

[RuudTom83]

Welcome to modern football...stats are far too overly used, it makes a lot of discussion boring.

We should just replace the points system...3 points to the team with the most possession and 1 point for the most xg

 

[Grylte]

You often see a player get MOTM after a bad game because he scored the goal in a 1-0.

 

[Sky1981]

I believe stats dont tell the whole story.

But i still believe that at one point statistic tells its own story.

When You score 10 beautiful goals a season Maradona like, stats don't do you justice

But if you score 30 goals a season for 15 season without fail all over Europe, the stats don't lie no matter how they claim it's an easy tap in.

At the end of the day goals is what matters in football

 

[adexkola]

Do you watch every game in the PL at full attention, being able to absorb all the events happening on the pitch in real time?

For those of us that can't (cause we have jobs and stuff), statistics provide an incomplete but the best picture of what actually happened, including how a player performed.

 

[sullydnl]

With the possible exception of defending (which can be tricky to measure statistically), I struggle to think of material positive contributions a player can make that wouldn't reflect themselves in the stats in some way. If a player is playing well regularly, that should largely be visible statistically. And if you think a player is/isn't playing well but the stats are at odds with your opinion, there's quit a high likelihood your opinion is where the fault lies. And of course when people say they've watched a player they've usually only watched a fraction of their games, whereas the stats cover the entirety.

All that said, it comes with the massive proviso that "visible in the stats" means visible when placed into context and with a high level of statistical understanding. Whereas not being statisticians or professional scouts, this is where the sort of amateur analysis we do on this forum runs into trouble. So a pinch of salt is definitely needed when stats are thrown around on this forum.

However, if you think a palyer played badly and someone counters it with X, Y and Z stats, you should have an actual rationale for why those numbers aren't capturing what you subjectively view as the reality. So if you say a player is lazy and the stats show he runs and presses a lot, you need to either accept he isn't lazy or square him being lazy with him still being able to hit those numbers. Not just dismiss the stats off-hand as if they don't capture something material.

 

[Sassy Colin]

Is this going to be another thread about what utter bollocks xG is?

Oh please let it be another thread about what utter bollocks xG is!

 

[Oranges038]

tl;dr version for the rest of us, please Pogue

Basically all these stats mean are should have scored more or should have conceded more. They've just put a number on them.

 

[Mr Pigeon]

Definitely true that stats can mislead in an individual game (although not if you dig into all the available stats, stuff like possession lost, possession regained, number of pressurew blah blah) but they’re obviously much more relevant/useful when you combine data from lots of games.

This. I love stats when they're used properly but not when folk try to mention individual game xG before they've even commented on the match. It makes me wonder how many folk just look at stats to decide if the game was good or not, rather than watch it.

 

[rimaldo]

i’d need to see the stats about judging players by stats before i could form an opinion.

 

[Olmer]

The rule of thumb is that you form an opinion based on extensive observation and then try to support it with stats. Stats shouldn't be the main factor in forming opinions.
Having said that, stats can be main factor in debunking opinions. The reason why is that it's the fastest way. Normally nobody wants to spend hours of their time just to debunk a random person on the internet. So we throw around some stats and see if that person can challenge them. Fast and simple.

I believe there are stats that truly elite players simply shouldn't have and citing those stats is enough to challenge opinions. The most basic example - a GOAT level striker shouldn't have a goalscoring ratio of 0.3 or 0.4 in Europe. Doesn't matter if his name is Ronaldo or Luis Suarez.

Random remark: it will be easier to judge contemporary players with stats 30 years into the future because we will have comparison with their successors. Right now we don't have this future yet. If we're gonna have 2+ Haalands every decade from now on, it will certainly affect our perception and judgement of the last years. All the goalscoring feats of yesterday will suddenly mean less and might be perceived as first signs of new footballing era where those numbers are normalized.

 

[Cassidy]

i’d need to see the stats about judging players by stats before i could form an opinion.

 

[Solius]

They can help but yeah you need to combine it with the eye test. They should both work in partnership, not against each other.

 

[Razvan]

You can judge players only by stats, it just depends what type of stats/the granularity of the stats you have. Goals/assists/passes is obviously too simplistic, but I feel like the data analytics field in football has advanced quite a bit in recent years and if you use the more advanced metrics you can get a really confident estimate of what type of player a certain player is by only looking at stats.

 

[Mr Pigeon]

i’d need to see the stats about judging players by stats before i could form an opinion.

 

[Infra-red]

They can help but yeah you need to combine it with the eye test. They should both work in partnership, not against each other.

Exactly. If you neglect either statistics or the eye test, then you are not getting a full picture of the player's capabilities, which is why clubs incorporate both aspects into their scouting operations.

 

[Gehrman]

How do stats hold up for a player like Biscuits in his prime? Not talking about G/A but all the others.

 

[Physiocrat]

Can someone link an example of a good use of stats in their proper context?

 

[kouroux]

It's very possible that a player will have played very well yet not have anything to show for it in the key stats metrics (at least the headline ones). Equally a player can easily have a poor game but pops up with a moment in the game that gets him something that reflects well on the stats. Over the last three or four years there seems to be an intolerance towards any suggestion that how a player has performed may have no correlation with the a small number of headline stats.

You get an assist = you've played well

You haven't had any key passes = you've played badly.

To the point where 95% of discussion around a player is now 'LOL you think he played well? Lowest key passes in the team" or LMAFO - yeah he played so badly, that's why he got an assist" - and that's the end of the discussion.

I've watched players I thought had stinkers who got an assist or a goal and players who haven't registered on any key stat metic who performed well.

Yet this is rarely ever accepted now. Everyone seems to watch a game and form an opinion on how well someone played based on the stats that come out of it. As if they can form no opinion watching the game.

Same principle is applied to describe one player as the best passer if he has most assists. Assists are a particular aspect of passing

 

[JPRouve]

There isn't a single statistic that is meant to be used in isolation, regardless of the topic.

 

[Gehrman]

Stats also dont tell you about the quality of goals and assists. All those memorably moments from someone like Cantona can't really measured by a number.

 

[Rossa]

According to fotmob, Bruno was the best player in the PL last season... https://www.fotmob.com/nb/leagues/47/stats/season/17664/players/rating/premier-league-players

Quite mental then that so many supporters want him out.

I think stats are fine, but most of us aren't statisticians, and we need to acknowledge that just as we aren't managers or players. We know quite little atop our very own peak of Mount Stupid. Therefore, the most annoying thing with using stats is that people think they've found an objective truth in the stat they are using. So the arguments that you made from watching the game make no relevance because your mate, who didn't watch the game, went to fotmob and read some stats that showed you were clearly in the wrong.

 

[Rossa]

Definitely true that stats can mislead in an individual game (although not if you dig into all the available stats, stuff like possession lost, possession regained, number of pressurew blah blah) but they’re obviously much more relevant/useful when you combine data from lots of games.

But how many football fans do you think are actually able to use stats to such an extent they can form an educated approach to understand data in such a way? I mean, statistics is a fairly difficult and overreaching subject in and of itself. Reading football stats, you realise just how arbitrary they can be. For instance, you can read stats from one source that give you the impression of a world beater, but the other source use different metrics, and thus the player was average at best.

Seriously, there are football fans all around the world discussing and evaluating players based on stats and Youtube compilations. They have hardle ever, if ever at all, watched said player live once, and by live I mean live stream/TV, whatever.

 

[Skills]

But how many football fans do you think are actually able to use stats to such an extent they can form an educated approach to understand data in such a way? I mean, statistics is a fairly difficult and overreaching subject in and of itself. Reading football stats, you realise just how arbitrary they can be. For instance, you can read stats from one source that give you the impression of a world beater, but the other source use different metrics, and thus the player was average at best.

Seriously, there are football fans all around the world discussing and evaluating players based on stats and Youtube compilations. They have hardle ever, if ever at all, watched said player live once, and by live I mean live stream/TV, whatever.

Is this sort of data that hard to understand? Most of the stats and the way they're presented are fairly elementary aren't they?

 

[Pogue Mahone]

According to fotmob, Bruno was the best player in the PL last season... https://www.fotmob.com/nb/leagues/47/stats/season/17664/players/rating/premier-league-players

Quite mental then that so many supporters want him out.

I think stats are fine, but most of us aren't statisticians, and we need to acknowledge that just as we aren't managers or players. We know quite little atop our very own peak of Mount Stupid. Therefore, the most annoying thing with using stats is that people think they've found an objective truth in the stat they are using. So the arguments that you made from watching the game make no relevance because your mate, who didn't watch the game, went to fotmob and read some stats that showed you were clearly in the wrong.

Absolutely. Although, to be clear, it’s the supporters who want him out that are mental. Not the stats.

 

[adexkola]

Is this sort of data that hard to understand? Most of the stats and the way they're presented are fairly elementary aren't they?

It's not hard at all.

People just suck at math in general.

 

[johnnyteutonic]

I spent 10 minutes this morning wrapping my head around PSxG-GA. 10 minutes I will never get back.

Put simply, the 'GA' part is simply how many goals were actually conceded.
The PsXG part is after the shot was taken, what was the average likelihood (or expectation) of it being a goal?
Basically it's a probability that takes into account the quality of shot after it has been taken e.g. was it a scuffed shot or did have much power?

So what you get is a picture of how many goals the keeper *should have been expected to concede* vs how many goals the keeper *actually conceded*.

If the number is positive, then the keeper is doing better than expected, if it's zero it's average and if it's negative, it's worse than expected.
Of course, the stat needs to be used on aggregate so it will tell you a keeper's net performance over the course of a large number of games.

 

[johnnyteutonic]

With the possible exception of defending (which can be tricky to measure statistically), I struggle to think of material positive contributions a player can make that wouldn't reflect themselves in the stats in some way. If a player is playing well regularly, that should largely be visible statistically. And if you think a player is/isn't playing well but the stats are at odds with your opinion, there's quit a high likelihood your opinion is where the fault lies. And of course when people say they've watched a player they've usually only watched a fraction of their games, whereas the stats cover the entirety.

All that said, it comes with the massive proviso that "visible in the stats" means visible when placed into context and with a high level of statistical understanding. Whereas not being statisticians or professional scouts, this is where the sort of amateur analysis we do on this forum runs into trouble. So a pinch of salt is definitely needed when stats are thrown around on this forum.

However, if you think a palyer played badly and someone counters it with X, Y and Z stats, you should have an actual rationale for why those numbers aren't capturing what you subjectively view as the reality. So if you say a player is lazy and the stats show he runs and presses a lot, you need to either accept he isn't lazy or square him being lazy with him still being able to hit those numbers. Not just dismiss the stats off-hand as if they don't capture something material.

On the whole I would agree, although, I personally think that we don't (yet) have a very useful way to quantify how good a keeper is at the 'ball-playing' part of keeping.
On fbref, they only track pass completion percentage for 'launched' kicks, but don't really have any metrics that are useful for measuring a keeper's overall ability on the ball.

You can look at the number of touches a keeper makes, which will generally correlate with how much possession a team has but I don't think it tells you the *quality* of the passes.

For example, how good is a keeper at beating the press and ensuring the pass they make isn't a hospital pass?
What about how good a keeper is at initiating attacks?
I would argue that this is a material thing that can be quantified but unless there are proprietary models that can do this, I haven't seen such a stat used effectively in publicly available models.
The eye test remains the best for this, in my opinion.