Do 'Expected' goals statistics suck?

[a mangy Meowth]

Not necessarily, it just gets tiresome when analytics and "5on5 xG/60" numbers get thrown around so frequently.

Plenty of dumb shit gets posted on here from a wide variety of angles including over-reliance on numbers

You know what gets tiresome for me? People not being able to understand the basic underpinnings of statistics, so they suggest to just throw all of them out

 

[Machinehead]

For me, the problem lies in thati t's just a series of tracked data sets that someone decides, on a whim, to assign a higher value to because they may or may not have a theory as to how goals are scored or stopped. The problem is that there is no conclusion to the data, just the data. Which is why it seems devoid of any context. At least that's how I see it.

Always willing to have a discussion but it has to start with why is it valued over any other statistic, and can you show any level of if/then causality attributable to said data point(s).

As far as why it's valued, good xGF% teams win the Cup a lot more than teams that aren't. And now we're seeing teams like the Panthers, who were positing elite xG metrics years before they were good. The predictive value is proven.

The models are not designed on a whim nor are they designed with lack of theory on how goals are scored. Despite the internet meme of a bunch of nerds who have never seen hockey, a vast amount of hockey knowledge goes into creating models and tracking events.

There's this character in Moneyball, Pete. Pete is the meme. He's this out of shape nerd from some fancy college and it's heavily implied he knows nothing about the game. That's not reality. The real guy that Pete is based on was in his second front office at that point (previously worked for the Cleveland Indians) and played baseball at a high level.

He did go to Harvard for economics, that part is true. The part they leave out is that he played for Harvard's baseball team and football team. That doesn't fit the nerd meme.

The analytics people care about hockey a lot and know a ton about it.

 

[shadow1]

No, it doesn't suck. But it's one advanced stat that requires a bit of data to have an impact.

Example: Foerster-Poehling-Konecny had a 42.8% xGF last game... 0.13 xGF, 0.18 xGA. Look at 10-20 games and you'll have useful data.

 

[Beerz]

Wait.. but Jfresh says

 

[Machinehead]

Wait.. but Jfresh says

Actually, that's a good example.

I knew JFresh before he was this guy with 127k followers. Nobody consumes more hockey than he does. If there's anything about the game that needs to be known, he knows it.

 

[Beerz]

They are all garbage if you are trying to look at individuals with them.

I think they are good if you want them to tell you about your team.

 

[Poppy Whoa Sonnet]

They fail to capture cross ice passes right before the shot, which is one of the biggest predictors I think of a goal going in. It leads to some bias and rewards certain players ...cough...timo meier...cough who take lots of shots against squared up goalies.

 

[Dr Pepper]

Plenty of dumb shit gets posted on here from a wide variety of angles including over-reliance on numbers

You know what gets tiresome for me? People not being able to understand the basic underpinnings of statistics, so they suggest to just throw all of them out

.....Again, not necessarily, nor did I say they should be "thrown out".

 

[a mangy Meowth]

.....Again, not necessarily, nor did I say they should be "thrown out".

What else could you possibly mean by suggesting we shouldn't even bother to keep data on this stuff?

 

[Dr Pepper]

What else could you possibly mean by suggesting we shouldn't even bother to keep data on this stuff?

Just because I personally don't see the point in it doesn't mean others can't enjoy it, which seems to be the case at least with a few.

 

[a mangy Meowth]

Just because I personally don't see the point in it doesn't mean others can't enjoy it, which seems to be the case at least with a few.

Doesn't jive with your attitude even one page ago. But all good

 

[GirardSpinorama]

xG in its current form is fundamnetally flawed and way too basic to be significantly useful for hockey. We use billions of parameters in the latest AI models to predict answers and these silly xG models use what? A few dozen?

 

[glenngineer]

IMO. I think using them in big sample sizes is good. For example, if you look at season to season xGF stats. Majority of the good teams are in the top half , with a few outliers. So it is largely doing a good job identifying good teams.

But using it in single game samples will give you some crazy results many times.

I don't disagree with you but go look at Nashville's advanced stats and it's a tale of two stories, stats versus reality. Our organization (GM, coach, broadcast) have all said if we were just a little more lucky with our bounces, we'd be knocking at the door of a wild card slot.

Having watched all the games this year, the xGF stat doesn't correlate to what I saw on the ice. This team could barely put three passes together consistently up until about 3 games ago and then all of a sudden, they looked like an NHL team.

I get this is an outlier over the course of a season but man, the eye test and the analytics did not match up at all.

 

[a mangy Meowth]

xG in its current form is fundamnetally flawed and way too basic to be significantly useful for hockey. We use billions of parameters in the latest AI models to predict answers and these silly xG models use what? A few dozen?

You're using one assumption here to canvas entire models that you state that you don't even understand. Doesn't that seem like a flawed analysis?

 

[GirardSpinorama]

You're using one assumption here to canvas entire models that you state that you don't even understand. Doesn't that seem like a flawed analysis?

xG models attempts to canvas a chaotic game like hockey based on flawed assumptions and clear lack of understanding of the game seems flawed to me.

 

[a mangy Meowth]

xG models attempts to canvas a chaotic game like hockey based on flawed assumptions and clear lack of understanding of the game seems flawed to me.

Oh for sure, but they aren't completely devoid of value like you're suggesting

 

[OrangePMD]

They fail to capture cross ice passes right before the shot, which is one of the biggest predictors I think of a goal going in. It leads to some bias and rewards certain players ...cough...timo meier...cough who take lots of shots against squared up goalies.

Yes, but they use previous events, time elapsed and distance from said previous event, which sort of works as a proxy for several situations, in long term. It's bit of a double edged sword, do you want more accurate description of the scoring chance, or do you prefer more data ?

If we take this Danforth goal in the op as example, there is no way to put into numbers how often Danforth scores from a tap-in, because Werenski turns a walking speed rush after an O-zone turnover to a backdoor tap-in, because it just doesn't happen that often. Where as, oz turnover, 9 seconds, wrister from 6 ft, random shooter happens quite often.

 

[Beezeral]

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.

This can't be serious. You are really going to argue that a goalie who stopped 21/22 easy shots had a better game than a goalie who stopped a bunch of high danger chances but gave up more goals?

 

[Poppy Whoa Sonnet]

Yes, but they use previous events, time elapsed and distance from said previous event, which sort of works as a proxy for several situations, in long term. It's bit of a double edged sword, do you want more accurate description of the scoring chance, or do you prefer more data ?

If we take this Danforth goal in the op as example, there is no way to put into numbers how often Danforth scores from a tap-in, because Werenski turns a walking speed rush after an O-zone turnover to a backdoor tap-in, because it just doesn't happen that often. Where as, oz turnover, 9 seconds, wrister from 6 ft, random shooter happens quite often.

The events in question are just shot attempts right? Like it works for rebounds, but not for "player A held onto the puck at this position of the ice for x seconds"

 

[Dr Pepper]

This can't be serious. You are really going to argue that a goalie who stopped 21/22 easy shots had a better game than a goalie who stopped a bunch of high danger chances but gave up more goals?

Statistically? Yes. Hard to argue otherwise, it's right there.

 

[OrangePMD]

The events in question are just shot attempts right? Like it works for rebounds, but not for "player A held onto the puck at this position of the ice for x seconds"

No, probably everything that gets logged as an event, shot attempts, turnovers, hits, faceoffs etc.

Those numbers and events from the Danforth goal are straight from the play-by-play log. Play By Play
And Moneypuck model is here. MoneyPuck.com -About and How it Works

 

[Bear of Bad News]

Statistically? Yes. Hard to argue otherwise, it's right there.

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?

 

[WarriorofTime]

I would look to it over the course of a long sample (i.e., 20 games, entire season) and not so much on a single game basis, as it's basically looking at where shot attempts come from and how often those lead to goals. Getting too in the weeds on an individual game will lead to a lot of messy looking stuff.

 

[VivaLasVegas]

I feel like everytime I go under the hood and really look at them, they just do not represent reality.

For example, Moneypuck had this Justin Danforth goal as .16 XG, which to my understanding is them saying this goal should only go in 16/100 times on an average NHL goalie (please correct me if I am wrong). IMO, it's closer to 96/100 than it is 16.




Idk who watched the Rangers/Jackets tonight but it's absolutely laughable that these stats 'say' the Rangers should have won that game most of the time. They were absolutely dreadful defensively and in pretty much all phases of the game.

I feel like I laugh my ass off at the 'Win O Meter' more than half the time they get posted in GDTs/PGT's (again as being wrong and unreflective of reality)

Help me. Am I misinterpreting the information? Should these stats and source (Moneypuck) be thrown out the window? Is there a better source (where you can actually pinpoint specific shots/goals and see their 'expectedness')

"Deserve To Win" is the last pathetic refuge of the fans of a losing team.

 

[Dr Pepper]

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?

Oh, 100%

That's like saying a skater doesn't deserve a Hart Trophy or Art Ross (or Rocket Richard), due to strength of schedule or calibre of goalie faced.

Still gotta do the work, and some are just better at it than others.