Wednesday, March 16, 2011

Goalie Point Shares

Hockey Reference introduced a new point shares metric recently. Is it a valuable stat that sheds more light on historical goalie performance?

Let's look at the all-time top 20 list:

1. Patrick Roy
2. Martin Brodeur
3. Terry Sawchuk
4. Tony Esposito
5. Glenn Hall
6. Jacques Plante
7. Curtis Joseph
8. Ed Belfour
9. Dominik Hasek
10. John Vanbiesbrouck
11. Roberto Luongo
12. Rogie Vachon
13. Sean Burke
14. Harry Lumley
15. Gump Worsley
16. Tomas Vokoun
17. Grant Fuhr
18. Tom Barrasso
19. Bernie Parent
20. Billy Smith

I conclude pretty confidently that the above list doesn't come close to passing the common sense test, as it deviates significantly from virtually every other ranking method out there. There's lots that I could nitpick, but I'll just point out that I don't think I've ever seen any other career list, statistical or subjective, that had Joseph, Belfour and Hasek ranked in that order.

It looks like the percentage of credit given to the goalies for team results is too high, and longevity has an excessive impact on the rankings. I probably rate Sean Burke higher than most, especially for his work in the second half of his career, but he still shouldn't be anywhere close to 13th all-time. In fact, given that Burke ranks 12th in career games played, he's a pretty good test case to see whether you are evaluating a goalie's actual performance or just giving them points for showing up for work. I can't help but conclude that Goalie Point Shares does far too much of the latter.

If you're looking for a one stat representation of a goalie's career, I'd still recommend sticking to era-adjusted save percentage or GVT.

14 comments:

Topher0820 said...

what if you rank by point shares per game?

Bettman's Nightmare said...

Yeah, I thought it a little fishy when Sean Burke led the league in goalie point shares in 2000-01 and Murray Bannerman in 1983-84.

The Contrarian Goaltender said...

I ran the per-game numbers for the top 100 goalies in GPS, and here's the top 20:

1. Ken Dryden, .248
2. Bill Durnan, .237
3. Tim Thomas, .231
4. Roberto Luongo, .220
5. Dominik Hasek, .213
5. Clint Benedict, .213
7. Jacques Plante, .207
8. Tomas Vokoun, .206
8. Bernie Parent, .206
10. Alec Connell, .205
11. George Hainsworth, .203
12. Tony Esposito, .202
13. Henrik Lundqvist, .199
14. Ryan Miller, .198
15. Glenn Hall, .196
16. Tiny Thompson, .194
17. Dave Kerr, .193
17. Patrick Roy, .193
19. Turk Broda, .192
20. Ilya Bryzgalov, .191

It's pretty obvious that there's not nearly enough of a spread in the results. For example, the gap from Jacques Plante to Ilya Bryzgalov is equivalent to just 4 wins per 500 games played.

If you look at GVT/game, Plante has an edge of 0.19 per game over Bryzgalov, which equates to 95 goals per 500 games played. That would be something like 16-19 wins per 500 games played, or about 3 games per full season. I think that's a much more reasonable estimate of the difference between the two goalies.

The correlation coefficient between career GPS and career GP is 0.92, which seems way too high to me. It looks like for the most part they are just splitting some arbitrary portion of the team result to the account of the goaltender, rather than figuring out how much credit he actually deserves for his performance.

Robert Vollman said...

First off, don't diss Murray Bannerman. When Tony Esposito had a 4.82 GAA, dude has a 3.38.

I kept getting Murray Bannerman cards as a kid. I think they sent every single one to my local shop.

Secondly, this statement is up for a Golden Elbow as the biggest understatement of the year:

The correlation coefficient between career GPS and career GP is 0.92, which seems way too high to me

P.S. I know Esposito was 40 at the time. And I know that random-nobody goalie Warren Skorodenski was 3.22 to Bannerman's 3.83 the next year, but still - don't diss the Bannerman. You wouldn't like him when he's angry.

R said...

When I take a look at this list, the onlygoalies that I think I would worry about going up against are Roy, Shawchuk and Luongo, the rest have flws that re easy to exploit.

overpass said...

Point Shares awards about 1/7 of team success to the goaltenders. This is not based on any empirical results, just the intuition of the stat's creator (who is not a subject matter expert in hockey, as far as I know - he's a sports-reference generalist.) See the following explanation.

* Why 7/12? At even strengh a team has six players on the ice, five skaters and one goalie. Imagine each of these players having two chips to contribute to one of two buckets: offense and defense. Collectively the skaters will contribute five chips to the offensive bucket and five chips to the defensive bucket. However, the goalie will contribute both of his chips to the defensive bucket, giving the defensive bucket seven of the twelve chips.

There is no support given for this assumption. One could equally as well assume that goaltenders only contribute on defence, and are one of six players who do so, so goaltenders should get about 1/12 of the credit for team success.

Gabriel Desjardins recently posted about the value of goaltenders in the NHL, and concluded goaltenders drive something like 8-10% of winning in the NHL That's a current estimate, and I don't know how well it holds up for all of NHL history. I don't think anyone does. But at least it's supported by data and not just the intuition of a stats guy who doesn't know hockey.

The point is that, if the metric is designed to give goaltenders in aggregate too much credit for team success, then the replacement level save percentage that the method uses, whether explicit or implicit, is set too low. Goalies get too much credit for showing up.

Anonymous said...

Lol Thomas sets the all time record for save percentage and you still haven't mentioned him once this year. Guess you're just butthurt over you butt buddy Hasek coming up short again. Stay jelly kid, your silence speaks volumes.

The Contrarian Goaltender said...

I'm a fan of Tim Thomas, I think he's largely underrated by many who want to give all the credit to the team in front of him, and yes he had a terrific season. You're seeing something that simply isn't there.

Cognition said...

I ran the numbers for the game situation sensitive GA30 formula you posted about last year, for every season since the lockout. You can check it out here if you're interested http://hockeycognition.wordpress.com/ga30/

Unknown said...

So, someone refuses to back down in an argument against me that wins are an important stat, and that save percentage isn't reliable as an indicator of goalie quality because of shot quality. They are specifically using Chris Osgood as an example because of his 400 wins. What do I say to shut him up?

I'm trying to find some examples of adjusted save percentage, but I'm sure that won't be enough for this idiot. Help me out?

Anonymous said...

Lol another prototypical performance from Re8oundo 7oun6oal tonight. He sure is "the best ever". How many times is he going to have to let in 5+ in the playoffs before you admit he is a huge choker? I wonder if his teammates think his great regular season stats make up for costing them another chance at the cup? LOL.

The Contrarian Goaltender said...

Durfee: You could try this argument.

Matt Kydd said...

Yeah, based on charisma and clutch performance and Stanley Cups and toughness, Billy Smith should be way higher than 20th. One of the best money goalies in history. And probably the toughest, which is an aspect that's important to team morale, as Timmy T showed in this years playoffs. Love your site, great stuff.
I also agree: don't diss Murray Bannerman. He had probably the best mask ever.

Hostpph said...

maybe he may not be in the best three players, but John Vanbiesbrouck is one of my favorites!