Advanced Stats Thread

[leafman101]

But adding those together, you're roughly just duplicating a GF/GA stat.
PP% + Fenwick% + SH% ~= GF
SH% + SV% ~= GA

It is still interesting to see, and might do a better job overall than simply GF/GA.

Sure to a certain extent, but isn't that what we are trying to do. Obviously GF/GA decides hockey games. But we are trying to delve deeper to see what causes a team to score more/give up less goals.

 

[worm]

Here is 5-5 GF/GA
1 CHICAGO
2 PITTSBURGH
3 NY RANGERS
4 BOSTON
5 ANAHEIM
6 MONTREAL
7 DETROIT
8 LOS ANGELES
9 TAMPA BAY
10 WASHINGTON
11 PHOENIX
12 TORONTO
13 ST LOUIS
14 COLUMBUS
15 OTTAWA
16 VANCOUVER

 

[zeke]

I still think the entire team stat approach is backwards from the start.

OFFENSE

- shots are simply not a good proxy for team offense. At all.
- offense is a matter of indi idual talent
- the only useful way to predict offense is to create a useful projection for each individual player and sum it all up in an intelligent way. My 2yr stats do this in a very basic way but its the right kind of idea.

DEFENSE

- most importantly, goalie performance should only be projected based on the goalies' own performance track record. Like offense, this is an individual stat not a team stat.

- shots against ARE a useful proxy for team defense. Not perfect by any means, but useful.
- adjusting shots against for quality might be useful, but i'm not sold on that
- still, i would use recent team shots against totals to project future team defense

- overall defense, then, would be projectrd based on recent team shots against totals and recent individual goalie save percentage


The more i think about it the more i realize that i think the biggest problem with the posession stats is that they treat shots for and against as having the same meaning and value, when i don't think they do.

Shots for are largely under your control, and its quite easy for talented players to pass up easy shots in favor of trying for better scoring chances, and thus beating the odds. Strategy can quite easily effect the quality distribution of shots for - some coaches quite clearly embrace the 'shoot
from everywhere and crash the crease" approach, others clearly embrace the opposite approach.

Shots against don't work the same way - it is very hard for any team to significantly effect the quality distribution of shots against compared to other teams, since every team has the same basic idea of trying to prevent shots against from in close.

I don't know for sure but i'd wager that shots for have a much weaker correlation with goals for than shots against have with goals against (though then again save percentage might **** with that a bit).

One thing i'm not convinced needs to be factored in separately is special teams. I think that would already be accounted for in the other stats.

So yeah, my projection model would be:

1.individual production projections totalled up for team offense, with shots being irrelevant
2.team shots against totals based on recent trends for team defense
3.individual save percentage projections factored into the shots against for team defense

As a sidenote, it's obviously much easier to evaluate the impact of offensive and goale roster changes than defensi e changes.

 

[hockeylover]

You know what's going to be really annoying? You don't need any advanced stats to tell you a team can't play at a 140+ point pace for very long. But we're really going to hear it from the metrics guys as soon as we lose a few. Regression and shit.

 

[zeke]

Interesting discussion going on on twitter the past coupla days, thanks to a leafs-supporting post from a respected sabermatrician a couple days ago.

The original post: http://blog.philbirnbaum.com/2013/10/corsi-shot-quality-and-toronto-maple.html

Corsi, shot quality, and the Toronto Maple Leafs

he Toronto Maple Leafs had a decent season in 2012-13, finishing fifth in the conference and making the playoffs for the first time since 2004. But, perhaps, we fans of God's Team shouldn't get too optimistic. For months now, hockey sabermetricians have been arguing that the Leafs were still a bad team -- a bad team that just happened to get exceptionally lucky.

But ... I've been fiddling a bit with the numbers, and I'm not sure I agree.

Before I get to my own case, though, let me tell you why the consensus says what it says. First, Sean McIndoe has an excellent Grantland article that summarizes the issue. Second, when you're done that, here's my own summary, which is a bit more statistical......

and a good response: http://www.sbnation.com/nhl/2013/10/16/4842102/nhl-stats-corsi-predictive-value-shot-quality

Is Corsi really as predictive as we think?

Most analysts rely heavily on shot differential, or Corsi, to make predictions for teams. When they do that, they're largely ignoring shooting percentages -- a decision they make based on evidence that shooting percentages have a lot of random variance, and that shot differential is actually a better predictor than the number of goals scored or wins compiled.

But Phil Birnbaum had an interesting post Tuesday challenging that idea, proposing that shooting a lot might be correlated with taking lower-quality shots.

He pointed to some pretty well-established score effects, or changes teams make depending on the score of a game, wherein teams adopt a defensive shell with a lead, taking fewer shots but for a higher percentage. He additionally noted that on some plays a player might pass on a low-percentage shot for the chance at a higher-percentage one, and wondered if there weren't strategies and/or random fluctuations that could result in both a low shot total and a high shooting percentage.....

and another response that is not quite as good: http://www.pensionplanpuppets.com/2013/10/16/4842716/rebutting-phil-birnbaum-on-shot-quality

Rebutting Phil Birnbaum On Shot Quality

I'm not a big believer in shot quality at the team level, as you may already know. But I'm not going to argue about shot quality today. I'm also not a big believer in the ability of 48 game sample sizes to settle team-level statistical debates, but I'm going to try and stay away from that angle too. Instead I'm going to assume that the central premise of Phil's argument is true: that a team could play in a defensive shell that lowered their Corsi but increased their SH% dramatically. Even if that's true, I think you still need to demonstrate that last year's Leafs team did in fact deploy that strategy with that result. Phil doesn't do that (not that I can see, anyway) but I think I can pretty conclusively argue that the Leafs didn't work that way.......

 

[BeLeafer]

It's probably been pointed out previously but the weakness of Corsi can be demonstrated by looking at the results for the Leafs under Maurice vs. Carlyle.

2006-07: +4.3, 49%
2007-08: +1.7, 44%
2008-09: +1.4, 41%

Interestingly, win percentage tracks remarkably well with shot differential for Maurice's teams. He was a major proponent of fire at will, quantity over quality.

However, when you compare this against another coach and system, the correlation breaks down. Here's Carlyle's nos. with the Leafs the last two seasons:

2012-13: -6.0, 54%
2013-14: -6.2, 86%*

This is a notably a small dataset (the use of the term "sample size" would be incorrect and its common misuse is a personal pet peeve of mine!). However, you can find the same trend with Carlyle's Anaheim teams.

2011-12: -1.0, 42%
2010-11: -3.8, 57%
2009-10: -3.2, 48%
2008-09: -0.2, 51%
2007-08: -0.7, 57%

His teams consistently show a negative shot differential and, if anything, it correlates negatively with win percentage (e.g. same results in 2011 and 2008 but vastly different shot differential). Clearly, Carlyle uses a coaching approach that focuses on shot quality rather than quantity, unlike the two coaches that preceded him in Toronto. The results belie to the Corsi hypothesis - no doubt created by a coach that thinks shot quantity improves winning chances.

 

[zeke]

well we can't push that too far - what the Leafs are doing under Carlyle is quite exceptional, and there is a very large amount of data which does indicate that, at the very least, these kind of exceptions are very rare, and maybe even unsustainable.

There's another descriptions that might be more apt: "The Leafs are a very talented team playing very badly, and winning just based on skill alone".

Of course, nobody in the media wants to actually formulate that argument, because then they would have to admit that the Leafs are playing badly.


But it is funny that when the Ron Wilson coached leafs had a top-5 Corsi/Fenwick and finished 29th in the league, nobody, and I mean nobody, suggested they were a team that was just underachieving and was bound to improve the next year simply due to regression to the mean.

 

[BeLeafer]

well we can't push that too far - what the Leafs are doing under Carlyle is quite exceptional, and there is a very large amount of data which does indicate that, at the very least, these kind of exceptions are very rare, and maybe even unsustainable.

This is why I went to Carlyle's numbers in Anaheim. For the coach, it's not exceptional. This is, as the numbers suggest, a coaching issue. Carlyle is clearly telling his teams that firing at will is not the way he likes to go.

In terms of their current record, it's obviously unsustainable. The goaltending is the difference between this 6-1 and a 3-3 or worse, imo. They aren't playing well at all. But they are executing on the offense well and it fits with Carlyle's general approach.

But it is funny that when the Ron Wilson coached leafs had a top-5 Corsi/Fenwick and finished 29th in the league, nobody, and I mean nobody, suggested they were a team that was just underachieving and was bound to improve the next year simply due to regression to the mean.

I did not know that! Just more grist for the advanced stats mill, I guess.

 

[zeke]

well we can't push that too far - what the Leafs are doing under Carlyle is quite exceptional, and there is a very large amount of data which does indicate that, at the very least, these kind of exceptions are very rare, and maybe even unsustainable.

There's another descriptions that might be more apt: "The Leafs are a very talented team playing very badly, and winning just based on skill alone".

Of course, nobody in the media wants to actually formulate that argument, because then they would have to admit that the Leafs are playing badly.


But it is funny that when the Ron Wilson coached leafs had a top-5 Corsi/Fenwick and finished 29th in the league, nobody, and I mean nobody, suggested they were a team that was just underachieving and was bound to improve the next year simply due to regression to the mean.

totally screwed up that sentence. Meant to end it with "admit that the Leafs are very talented".

 

[BeLeafer]

Funny, even so, I'd read it that way.

 

[PlayerToBeNamedLater]

The Leafs don't play very good team defense. Part of it is they play a more open game compared to a lot of teams, part of it is a lack of possession, part of it is personnel.

The Leafs very much rely on their elite offensive talent and elite goaltending to win them games, which isn't a bad thing at all. It's actually a good thing, because you can teach team defense, you can't teach goal scoring.

The Leafs overall shooting % the last 3 years: 12.3% (2nd), 11.5% (1st), 9.8% (6th). This is a stretch of 140 games of elite shooting %. At some point, you just have to admit that they are the exception to the rule.

 

[PlayerToBeNamedLater]

Sure to a certain extent, but isn't that what we are trying to do. Obviously GF/GA decides hockey games. But we are trying to delve deeper to see what causes a team to score more/give up less goals.

If you create some forumula that prorates a combination of goals/corsi for every position player, sv% for goalies for every roster based on 100-150ish game sample, maybe with some exceptions that factor in decline for older players, growth for younger players, I think you'd get a very accurate picture of the quality of each team.

 

[Montana]

I know you would.

 

[PlayerToBeNamedLater]

A good book is "The Numbers Game: Why Everything You Know About Soccer Is Wrong"

Not about hockey, but there are some intersting numbers and concepts that could be applied to hockey because both games are similar in a sense that both aren't a series of independent events, but continuous play, which is a lot harder to breakdown.

One interesting point made was that nowadays everyone is focusing on possession and how to score goals (in both hockey and soccer), but the best approach to counter this would be to focus on how to prevent them. It's not about scoring a lot of goals, it's just about outscoring your opponent.

 

[hockeylover]

Pretty close to just abandoning the advanced stats debate altogether. It's pretty exhausting. If I hear regression to the mean one more time... some non-Leaf fans can't even acknowledge the Leafs have a good scoring team.

 

[MindzEye]

It's really not worth it. When the Leafs don't suck this year (and I'm pretty much convinced of that now that we're well into the season, have looked terrible imo, yet manage to out talent/skill most teams by such a wide margin we're winning games), they'll just scream "outlier" and we're going to play this same game with them again.

 

[Montana]

I dropped it a long time ago, so hearing people complain about the debate has become more annoying than anything, as I don't follow anyone that brings up the other side of the debate.

It's almost like hearing about the Westborro Baptist Church.....on they're own they're completely irrelevant and I'd never hear about them, if it wasn't for people feeding their hype machine, by complaining about them all the time.


The more people ignore the nonsense, the better off everyone is......in the end, the truth has a way of sticking up for itself.

 

[Bleedsblue&white]

Enjoying this immensely, glad to see there is good debate going on.

Just for fun here:

The Leafs could be an outlier. Not their play last season...The Team itself, coached a certain way by a certain coach. I mean that in every example ("7 out of 10 teams that win game one win...")there are the exceptions. We say luck, randomness, whatever. I am saying that maybe the numbers ARE predictive of wins, but in no way a guarantee, and we are the team(s) that really do not make "sense," in the Corsi world.

We like to break things into seasons, but with Carlyle coaching, we need to break this team into...generations maybe? Iterations?
As long as Randy is coaching this style, and the team signs the players it does, we will continue to fall outside the norm, and win doing it.

 

[zeke]

It's really not worth it. When the Leafs don't suck this year (and I'm pretty much convinced of that now that we're well into the season, have looked terrible imo, yet manage to out talent/skill most teams by such a wide margin we're winning games), they'll just scream "outlier" and we're going to play this same game with them again.

It's true. We have to realize there's literally no sample size big enough for them to ever consider non-corsi/fenwick success as legit.

 

[MindzEye]

On a related note, does anyone actually know how to pull those shot location charts off of NHL.com's icetracker? I had some tracking ideas I wanted to play around with but can't manage to find the shot charts.