Do 'Expected' goals statistics suck?

[Breakers]

LOLOLOL

A goalie with a .720 save percentage tends to do that to models

 

[Midnight Judges]

Imo

Hockey prediction is closer to pinball than baseball. Much closer.

Statistical comparison and probabilistic outcomes are not something many of our brains like but the puck is either on edge or flat when the one timer comes through, the goalie perhaps can’t even see that an opponent is shooting, the puck hits a guy in the pants, then a skate and goes in 5 hole.

This is pure chaos.

Compare this to: guy throws his 52nd fastball and guy swings.

One is way easier to model than the other, much more predicable with results that are way easier to measure even in weird ways.

What advanced stats do and are is just a handful of measuring sticks. They attempt to place the same measuring stick on every team, player, and event. They can’t help their job is to define chaos.

The root of your problem is ice hockeys chaotic nature, not the evaluation of it.

Understanding their fundamental source will allow you to recognize them for what they are.

I’d suggest that they’re a great tool that joins others to help describe historic events and predict future ones.

Uh pinball is a legit skill where the best players consistently win.

 

[Romang67]

Very true. It’s a fun thought though!

That combined with pre-shot puck movement would almost certainly be the best way to evaluate goaltenders and goal scorers, so it would absolutely be super fun! Just really difficult.

 

[Bear of Bad News]

You're going to keep pretending save % gives any weight whatsoever to the difficulty of shots faced? By all means, have at it.

No, in fact, I'm saying exactly the opposite.

 

[StlBigFly]

Uh pinball is a legit skill where the best players consistently win.

Just like ice hockey.

You can get a predicted outcome. We expect McDavid to make the playoffs, lead the league in scoring, try for a cup and we get that. Whether a shot from a specific spot is going to go in or not is just…it is what it is. It’s what we can measure.

I feel there is just an inherent amount of randomness in hockey that makes things fuzzy when you look from certain perspectives, like shot quality. You can win when you aren’t doing the right things and you can lose when you are doing them. Things like golf and baseball are way less like this.

My best is fish tales. Thousands of games. Really good at that one. It’s really satisfying when you get a multi ball and get it under control, abuse the little boat in the middle to infinity and beyond. It’s like you’re a juggler.

 

[a mangy Meowth]

You're going to keep pretending save % gives any weight whatsoever to the difficulty of shots faced? By all means, have at it.

But I think I'd rather take a goalie with a .930 sv % over another goalie at .902, even if he faces "tougher shots".

Lmfao, you are all twisted up!

 

[Dr Pepper]

No, in fact, I'm saying exactly the opposite.

What?

You said "You're seriously advocating that we treat all shots in a game as equal and that we should ignore the difficulty that each goaltender faces?", which is exactly how save percentage is calculated.

I replied that a goalie's actual save percentage is far more valuable than "expected shot difficulty" or however you want to call it.

What am I missing here.

Lmfao, you are all twisted up!

Just trying to follow the argument.

 

[a mangy Meowth]

What am I missing here.

Apparently literally 100% of it

 

[Bear of Bad News]

What?

You said "You're seriously advocating that we treat all shots in a game as equal and that we should ignore the difficulty that each goaltender faces?", which is exactly how save percentage is calculated.

I replied that a goalie's actual save percentage is far more valuable than "expected shot difficulty" or however you want to call it.

What am I missing here.



Just trying to follow the argument.

This isn't difficult. You're advocating for using save percentage as-is. I asked if "you're seriously advocating..." those statements above. That's why I said "you" (and not "me").

Let me boil down my opinion since you seem to need that,

• Not all shots in a game are equal.
• We should not ignore the difficulty that each goaltender faces.
• There are better ways of doing things than save percentage.

What is your opinion?

 

[Bear of Bad News]

Apparently literally 100% of it

See, it all comes back to average shot difficulty.

 

[quackquackquack]

It might just be my "biased" eye test but I don't always feel like the data matches up with the game. For instance, data says Ducks are being pummeled and the goalies are saving our ass (still happens a lot) but the eye test is that there are a ton of outside shots or shots where there isn't a lot of traffic and goalie picks up shooter right away.

 

[Machinehead]

It might just be my "biased" eye test but I don't always feel like the data matches up with the game. For instance, data says Ducks are being pummeled and the goalies are saving our ass (still happens a lot) but the eye test is that there are a ton of outside shots or shots where there isn't a lot of traffic and goalie picks up shooter right away.

The thing about giving up a lot of outside shots is that the other team still has the puck and are the ones dictating play.

 

[quackquackquack]

The thing about giving up a lot of outside shots is that the other team still has the puck and are the ones dictating play.

sure but not dangerous by any means.

 

[Machinehead]

sure but not dangerous by any means.

It's always dangerous not having the puck.

 

[Voight]

I pretty much hate all the"advanced statistics". They're absolutely meaningless. Way too many variables in hockey.

Agreed. We dont need them.

 

[JKG33]

They are total junk in their current form. A 4th line plug shoveling three successive short range shots directly into a goalie who has the post sealed off at the side of the net is worth something like 10x the xGs of an open Ovechkin cross ice one timer. There's too much noise in a flow game like hockey to have the gall to claim to know how many goals you can expect to score from math.

That isn't to say they're not worth recording or considering, or that the concept isn't worth refining, but they need to be taken as part of a picture. And that part should be around 20% or so IMO.

And this kids, is why Connor Hellebuyck has an absurdly high GSAX or whatever abbreviation the spreadsheet jockeys are using

 

[OrangePMD]

It might just be my "biased" eye test but I don't always feel like the data matches up with the game. For instance, data says Ducks are being pummeled and the goalies are saving our ass (still happens a lot) but the eye test is that there are a ton of outside shots or shots where there isn't a lot of traffic and goalie picks up shooter right away.

It happens. I usually look up the stats after the game, sometimes I try to guess what they look like. And sometimes I'm way off. And sometimes I care to take a closer look at why would I had seen it so wrong. And sometimes I just write them off as dumb fancy stats and move on.

 

[ijuka]

They’re trying to galaxy brain that non-sense

Expected goals are even simpler than that. You expect to score 1 goal on every 10 shots on goal.

32 shots on goal? Expected goals is 3. Less than that and you got goalied, over that and the goalie played like shit.

Done

Shots on goal as a predictor is something like 0.10, while expected goals are something like 0.17. If you don't see the value, then that's on you.

Note I think the stat is far from perfect, but just because you use expected goals doesn't mean you can't use shots on goal, too. You obviously want to use all the predictors you can, and together they add up to the ultimate impression. You don't have to choose one over another.

 

[ijuka]

It might just be my "biased" eye test but I don't always feel like the data matches up with the game. For instance, data says Ducks are being pummeled and the goalies are saving our ass (still happens a lot) but the eye test is that there are a ton of outside shots or shots where there isn't a lot of traffic and goalie picks up shooter right away.

The idea is not to accurately represent the events of every individual game. The idea is that over 50 games or so, the overall impression averages to a pretty reasonable result, when the individual errors in expected goals' assumptions form a distribution around the truth.

Using individual matches completely misses the point. Did you know that with small samples such as the past 10 games, corsi actually outperforms expected goals? Expected goals requires a solid sample size, like 50 matches for a team, before it's truly useful.

 

[Machinehead]

They are total junk in their current form. A 4th line plug shoveling three successive short range shots directly into a goalie who has the post sealed off at the side of the net is worth something like 10x the xGs of an open Ovechkin cross ice one timer. There's too much noise in a flow game like hockey to have the gall to claim to know how many goals you can expect to score from math.

That isn't to say they're not worth recording or considering, or that the concept isn't worth refining, but they need to be taken as part of a picture. And that part should be around 20% or so IMO.

I always love the example of some plug taking a bunch of shots in the crease because like, when does that ever happen??

Sam Carrick, who is a damn good 4th liner, has 64 shots in 64 games. David Pastrnak has 272.

The idea that the elite players in this league, who generate quality looks, aren't also the guys generating most of the volume is completely contrary to reality.

 

[Bear of Bad News]

People are also quite good at identifying boundary cases with these sorts of metrics (and yes, Chris Dingman taking a shot from the slot won't have the same chance of scoring as Alexander Ovechkin).

Yet they seem to ignore (or say "it doesn't matter") with boundary cases - of which there are far more - on mainstream statistics.

(For evidence, see the "we can assume all shots are of the same quality" discussion earlier in this thread.)

 

[Machinehead]

People are also quite good at identifying boundary cases with these sorts of metrics (and yes, Chris Dingman taking a shot from the slot won't have the same chance of scoring as Alexander Ovechkin).

Yet they seem to ignore (or say "it doesn't matter") with boundary cases - of which there are far more - on mainstream statistics.

(For evidence, see the "we can assume all shots are of the same quality" discussion earlier in this thread.)

Also, it's always a thread when analytics produce a bit of a weird result like this, but there's never any flowers going the other way.

"Analytics never had the 2023 Bruins as the best team."
"Corsi told you Patrik Laine sucked ten years ago."
"We knew the Rangers wouldn't make it [enter season here]."
"Underlying numbers show Gustav Forsling was elite as early as 2020."

Those are never threads.

 

[Oilslick941611]

It kind of is though, isn't it?

If a goalie enters the game with a .914 save percentage on the season, then the "expected goals" should be 8.6% of the total shot attempts, should it not?



Still the first one, obviously, and hopefully when it comes time to figure out who wins the Vezina we don't have voters sitting around a table ignoring real stats and trying to figure out which goalie stopped the most "premium chances" instead.

No. Because it assumes everyone is playing at thier best and that every game and every shot in the same quality.

I don’t believe any of the major fancy stats predict anything other than a static shot at specific time frame. The game isn’t played on a spreadsheet and multiple events and instances can drastically change the outcome of a game. It’s dynamic, not linear. Baseball lends itself to fancy stars predicting the outcome, hockey not so much. Fancy stats in hockey should be used to help influence how managers and coaches think, not be the only source of that thinking.

Seriously what a player had breakfast can affect the game and no fancy stat can predict that.

Back to expected goals, it’s a bad stat IMO that should be taken with a grain of salt and not relied on for fans to say “we should have scored 3 goals that period”. It doesn’t take into account things that affect goals against like a bad line change, out of position D man, or a terrible forecheck that the D capitalized on for a breakaway, and once sgain assumes everything was equal and players playing at their best which doesn’t happen night in and night out. So it’s not a predicter of anything but what’s already happened and the box score already does that.

 

[Beerz]

The secret is to scream angrily at Google maps when I'm 10 minutes late despite leaving early enough that I should only have been 5 minutes late!

I'm sure that method translates to expected goals models too.

I just use the already available data from the scouts and listen to the radio or TV traffic guys.

 

[GeeoffBrown]

if you think about it, 0.16 is well above average, assuming a 0.900 save%, then 0.1 would be your average shot.

NHL goalies are very good and they do frequently stop cross-crease attempts