Sunday, November 2, 2008

Expected Shutouts, Part 2

Since we have already looked at goalies who did well by the expected shutouts measure, let's continue the discussion by looking at who didn't do so well. I have put together a list of goalies who have fewer career shutouts than expected (min. 30,000 minutes played):

GoalieSOExp SODiff%
Tommy Salo3737.1-0.10%
Chris Osgood4747.6-0.6-1%
Don Beaupre1717.6-0.6-3%
Felix Potvin3233.3-1.3-4%
Kirk McLean2223.3-1.3-6%
Ron Hextall2326.0-3.0-12%
Mike Richter2428.0-4.0-14%
Grant Fuhr2529.2-4.2-14%
Olaf Kolzig3541.2-6.2-15%
Bill Ranford1519.9-4.9-25%
Mike Vernon2738.2-11.2-29%
Ken Wregget915.4-6.4-42%

I saw Chris Osgood described recently on an Internet message board as "the definition of a league average goalie", which I think is a pretty apt description. Grant Fuhr once again puts up pretty unimpressive stats. I'm surprised Mike Richter and Olaf Kolzig rank so low, but not surprised at all by Bill Ranford or Mike Vernon.

Current goalies who have done poorly so far by this metric include: DiPietro (-5%), Roloson (-6%), Gerber (-16%), Fernandez (-18%), and especially Miller (-39%!).

Finally, let's look at which goalies have the most expected shutouts. That is, if you substituted a league average goalie in to play the same minutes, how many shutouts could we expect them to have? This table shows why the all-time shutout record is so hard to break - it requires durability, good teammates, and playing in a defensive era. Only two goalies in the post-WHA era had even an outside shot at breaking Sawchuk's record, and this list shows how the guy who eventually is going to do it had quite a bit of an advantage over the rest of his peers:

RankGoalieExp SO
1.Martin Brodeur77.6
2.Ed Belfour60.2
3.Patrick Roy47.7
4.Chris Osgood47.6
5.Curtis Joseph44.0
6.Dominik Hasek43.1
7.Olaf Kolzig41.2
8.Mike Vernon38.2
9.Tommy Salo37.1
10.Nikolai Khabibulin35.6


overpass said...

Rank Goalie Exp SO
1. Martin Brodeur 77.6
2. Ed Belfour 60.2
3. Patrick Roy 47.7

And there's the payoff! I wasn't sure where you were going with this series, but it makes it pretty clear that Brodeur had a big edge over his contemporaries in opportunities for shutouts.

Obviously facing fewer shots per game makes it much easier to post shutouts. However, there seems to be some evidence that Brodeur is responsible, at least in part, for some shot prevention. Your previous post suggested he might prevent 1 shot per game on average.

How much would facing an extra shot per game affect his expected shutouts? I'm sure it takes more time than it's worth to adjust for every factor when calculating a junk stat like shutouts, but I'd be curious to see if for nothing else than sensitivity analysis.

Anonymous said...

1 less shot per game... about 70 less shots in 70 games... which is about 2.5 games worth of shots, albeit spread out over a whole season...

... so 1 less shot per game probably works out to an extra shutout every 5+ yrs! Very minimal effect, I would think.

The Contrarian Goaltender said...

Brodeur's career average shots per game is 25.4, and the league average during his career has been about .905. That gives an expected career shutout total of 76 (doesn't exactly match the one above because I didn't do it season-by-season, but close enough). If we assume that an average goalie would face 26.4 shots per game, then the expected shutout total would drop to 69.

Note that to be fair we would have to adjust all the other goalies' numbers too. For example, based on comparing their numbers to their backup goalies over their careers, I'd estimate Ed Belfour to be even better than Martin Brodeur at shot prevention.

overpass said...

Thanks for the further analysis, CG.

I have to disagree with anonymous above - I don't think it's a minimal effect at all. Subtract 7 expected shutouts from Brodeur's total in the post below this one, and he's right beside Roy in shutouts over expected shutouts.

It's too bad, from an analytical perspective, that Brodeur has been so durable over his career. The small sample sizes of his backups make it difficult to find a shot-prevention effect with any precision - let alone looking to see if it has changed over time.

You never like to see anyone injured, and I hope Brodeur comes back quickly and fully recovered, but it will be interesting to see how other goalies will do with this Devils team.

My guess? The replacement goalie puts up good numbers, but worse than Brodeur's, and the Devils still finish with home-ice advantage in the first round, on the strength of their terrific forward corps. With Brodeur back for the playoffs, the Devils should still be in good shape.

Bruce said...

It's too bad, from an analytical perspective, that Brodeur has been so durable over his career.

What's really too bad is that the injury is serious. I feared as much from the video. The NHL will be without one of its brightest stars for the next 3-4 months.

It does however make for an interesting test case for Jersey backups and replacement goalies over a long period rather than carefully chosen appearances once a month or so.

My own prediction -- for what that's worth -- is that the Devils will be hard-pressed to make the playoffs. But as always, time will tell.

Anonymous said...

Brodeur to be out 3-4 months with injury. Looks like the sample size for Brodeurs backups will get a little bigger. We'll see if those theories of yours prove out.

Anonymous said...

7 more expected shutouts throughout 978+ career regular season games? That's a very small increase in frequency (+ 0.7%)... about 1 more shutout every 2.2 years.

No doubt the usual below-avg NJ backup goalie is going to look mediocre while playing in Brodeur's place while he recovers...

overpass said...

7 more expected shutouts throughout 978+ career regular season games? That's a very small increase in frequency (+ 0.7%)... about 1 more shutout every 2.2 years.

Shutouts are themselves a low frequency event. Look at it this way - if you assume that an average goalie faces 26.4 SOG/G instead of 25.4 SOG/G in Brodeur's minutes, the expected shutouts drop by 10%. Brodeur's shutouts over expected shutouts would go from 24% more to 35% more, or from JS Giguere territory to Patrick Roy territory. (Of course, no amount of adjustments would get him into Hasek territory.)

Now if you want to argue that shutouts aren't that important in rating goaltenders, I won't disagree. I'm just pointing out that the model is fairly sensitive to the assumption of "no shot prevention effect by goaltenders".

The Contrarian Goaltender said...

The model is fairly sensitive to a lot of things. I estimate, for example, the normal range of shot quality allowed by teams to be about three times the effect of normal goalie shot prevention in terms of goals against, which obviously has an impact on shutouts. There is the additional factor as well of someone who faces fewer shots to begin with, like Brodeur, facing a lot more "gimme" shutouts (e.g. 12-16 shots against, games which have an estimated 20-30% probability of a shutout). So the expected SO numbers for goalies like Brodeur may be lower than they would be if I calculated it on a game-by-game basis.

Shutout frequency is also very low, especially for the 1980s. For some of those guys a couple of lucky games could make the difference between an excellent and average showing based on this measure.

That is why I don't think we can make conclusions even when one guy is at 24% and another is at 38%. I think only extreme results are really telling - i.e. Hasek and Luongo are outstanding goalies, Bester and Froese were very much overlooked, and Vernon is colossally overrated.