Modern Football Fans & Stats

[TheBigIguana]

Yeah, I agree with the assertion that the stats the clubs are working with are probably a totally different level. Sample size though is a problem, especially for specific use-cases and positions. How much can you tell from a striker that gets 2-3 chances a game? That he should get more? Isn't that dependent on his midfield? 2-3 chances a game over 38 games is an entirely different sample size level than 20 chances a game over 82 game season, or 4 chances a game for 162 game seasons.

I also feel like it's a matter of time before they get it worked out. I'm just not on the stats bandwagon currently because the public stuff is very Meh. It's *better* than just counting goals and assists, I guess, but it certainly doesn't tell as much as the eyes. Yet.

They attempt to evaluate the effectiveness of the two to three chances by comparing them to similar chances and how effective they were across a large dataset. It's not a perfect solution but it does give you a number more useful than saying has 3 chances created per 90 minutes or something. They've got some pretty good metrics for attacking at this point tbh. Whether they'll prove over time to be effective will be interesting but they look solid now.

The problem is they have very limited understanding of how to quantify things like passing that doesn't lead directly to chances and especially defending. That's why Gomes could still be a good signing despite some doubts over his impact going forward. Because the public doesnt really have access to anything that shows what kind of impact his distribution actually has on our ability to score. That's where this stuff needs to improve.

 

[Prevenger17]

Isn't that dependent on his midfield?

It's also dependent on his own positional play, the areas he likes to attack compared to the areas his midfielders like to lash it into as well as a whole host of other related factors. The amount of stats, unseen stats and comparative data you would need to compile to make stats useful are only available to the Optas of this world, and only incredibly qualified, analytical people would be able to interpret them in a way that could be practically applied. Amateur spreadsheet merchants and the xG brigade get a crumb of data and think they can feed the whole world

 

[SerenityNigh]

I like stats but please don't try and tell me I should try and change my view on a player that I've actually watched because some tithead has made up an algorithm that is based on not watching the game.

This is a rather naive view of stats, to be fair. Everyone knows the talented xMeffs use their eyes and can even sometimes form complete sentences to support their favorite stats.

 

[RaleighBlue]

They attempt to evaluate the effectiveness of the two to three chances by comparing them to similar chances and how effective they were across a large dataset. It's not a perfect solution but it does give you a number more useful than saying has 3 chances created per 90 minutes or something. They've got some pretty good metrics for attacking at this point tbh. Whether they'll prove over time to be effective will be interesting but they look solid now.

The problem is they have very limited understanding of how to quantify things like passing that doesn't lead directly to chances and especially defending. That's why Gomes could still be a good signing despite some doubts over his impact going forward. Because the public doesnt really have access to anything that shows what kind of impact his distribution actually has on our ability to score. That's where this stuff needs to improve.

I realize the dataset problem, but that still gives a large picture rather than specific insight into an individual. You can say out of 300 strikers that got 3 chances of this quality a game, Player X performs poorly. But how many chances has Player X had? 90? That's a tiny sample size for an entire season. And could easily be a product of sheer statistical anomaly *for that instance*. It could take multiple seasons of that production for it to be actually statistically meaningful on an individual level.

It's also a problem of variability.

Even American sports struggle to effectively translate stats between leagues. It's much less of a problem over here, of course, because the transition is college -> pro league. But once in the pro league, you have the same opposition, the same sporting 'culture', the same 'styles of play'.

Football stats have to account for all of those things when trying to compare players across leagues, especially. How a player fares in Spain, or the Netherlands is not necessarily something that translates to England.

Basically my argument is: Football stats - not there yet; long way to go; lots of variables that need accounting for.

 

[SerenityNigh]

Yeah, I agree with the assertion that the stats the clubs are working with are probably a totally different level. Sample size though is a problem, especially for specific use-cases and positions. How much can you tell from a striker that gets 2-3 chances a game? That he should get more? Isn't that dependent on his midfield? 2-3 chances a game over 38 games is an entirely different sample size level than 20 chances a game over 82 game season, or 4 chances a game for 162 game seasons.

I also feel like it's a matter of time before they get it worked out. I'm just not on the stats bandwagon currently because the public stuff is very Meh. It's *better* than just counting goals and assists, I guess, but it certainly doesn't tell as much as the eyes. Yet.

the good stats will emerge eventually, but they'll be very different from what we're used to seeing. maybe we'll see some kind of "impact score" like baseball's WAR to measure what a player does in the game, but like you said above, most of it now is a very rudimentary approach because nobody knows what they're measuring.

 

[Blue Juice]

There are no stats for blammos, putting people on the weights or moment of magic that you can actually ENJOY.

Statistics take all of the enjoyment out of just watching the game, especially weird for Evertonians when we don't have that much to enjoy to begin with.

I couldnt care less that Bernards xSlip is over 100 or Gomes xDynamicPass is less than Real Madrids youth goalkeeper. I enjoy watching them because they're objectively good at the togger

 

[TheBigIguana]

I realize the dataset problem, but that still gives a large picture rather than specific insight into an individual. You can say out of 300 strikers that got 3 chances of this quality a game, Player X performs poorly. But how many chances has Player X had? 90? That's a tiny sample size for an entire season. And could easily be a product of sheer statistical anomaly *for that instance*. It could take multiple seasons of that production for it to be actually statistically meaningful on an individual level.

It's also a problem of variability.

Even American sports struggle to effectively translate stats between leagues. It's much less of a problem over here, of course, because the transition is college -> pro league. But once in the pro league, you have the same opposition, the same sporting 'culture', the same 'styles of play'.

Football stats have to account for all of those things when trying to compare players across leagues, especially. How a player fares in Spain, or the Netherlands is not necessarily something that translates to England.

Basically my argument is: Football stats - not there yet; long way to go; lots of variables that need accounting for.

Most of what I've seen doesn't really look to evaluate whether the ball actually goes in or not since finishing has a large amount of luck involved and rather try to quantify how much the player should score from the shots they take. If someone is a terrible finisher then the stat becomes pointless but often pro level forwards tend to finish at similar rates and the quality and quantity of chances tends to separate them. So after a season of 38 games there is a pretty good sample of performances to say x player should score y amount per 90 minutes.

 

[Billysgingham.]

Use your eyes. It's how I identified Lukaku as a guileless donkey years before the xG merchants even started to realise it.

The stats don't back it though,the stats say the opposite.

 

[Prevenger17]

Stats are useful to statisticians. The rest of us should watch football and say things like "that guy's crap" or "that guy's got a great turn of pace"

 

[Viennablue]

Stats can be used for certain aspects of the game .
But to base your value of a player solely on them is ridiculous. Like an assist might be a 2 yard pass across an empty goal but the pass to play in the assister makes it.
Or a winger puts in 5 world class balls that are not scored . Not his fault. But wont show up on his assists chart.

Take your point but...

That’s where the “key passes” and “chances created” stats come in.

Your absolutely right that goals and assists stats don’t mean everything. However, there are many other stats to look at when judging a player.

The people who say “watching a player is the best way to judge him” have a point. But you wouldn’t just watch them once. Is Steven Naismith as good as the hat trick against Chelsea suggests? No. You’d have to watch him 5-10 times to see how good he really is.

Realistically stats don’t show the whole picture. Neither does watching a player play 1 game.

A mixture of stats and watching a player is the best way to judge them. That’s what most football clubs do before spending their money.

The player’s lifestyle and potential off the field problems are also important. As is the player’s family situation, agent and language skills.

However, offer me the chance to watch a player once or look at 3 years worth of stats for a player efc might sign and if I was making the decision I’d take the stats.

 

[Prevenger17]

The stats don't back it though,the stats say the opposite.

Exactly, which is why you can't trust stats. Here's some interesting Rom numbers

Trophies Lukaku has won: 0
Amount of times Lukaku has hidden behind a centre half for the duration of a game he's not feeling up to: 1,000,000

 

[roydo]

However, offer me the chance to watch a player once or look at 3 years worth of stats for a player etc might sign and if I was making the decision I’d take the stats.

Thats like saying I would rather sign a player after watching him for 30 games rather than looking at 1 stat.

 

[Bluebonic]

People who bandy statistics about in contradiction of the evidence on view to those with functioning eyesight are , in my humble opinion, promoting witchcraft and should be burnt at the stake, or at least pilloried!

 

[SerenityNigh]

Take your point but...

That’s where the “key passes” and “chances created” stats come in.

Your absolutely right that goals and assists stats don’t mean everything. However, there are many other stats to look at when judging a player.

The people who say “watching a player is the best way to judge him” have a point. But you wouldn’t just watch them once. Is Steven Naismith as good as the hat trick against Chelsea suggests? No. You’d have to watch him 5-10 times to see how good he really is.

Realistically stats don’t show the whole picture. Neither does watching a player play 1 game.

A mixture of stats and watching a player is the best way to judge them. That’s what most football clubs do before spending their money.

The player’s lifestyle and potential off the field problems are also important. As is the player’s family situation, agent and language skills.

However, offer me the chance to watch a player once or look at 3 years worth of stats for a player efc might sign and if I was making the decision I’d take the stats.

I don't think anyone is saying sample size isn't an issue. I mean, look at Bungle, he's watching tons of webcams to come to his conclusions, not just once here or there. The police data backs up these observations.

 

[SerenityNigh]

Stats are useful to statisticians. The rest of us should watch football and say things like "that guy's crap" or "that guy's got a great turn of pace"

Yes, but is he quick?