Friday, November 7, 2008

Brodeur, Roy, and the Career Wins Record

The question was recently raised by someone in the Yahoo Hockey Analysis Group about whether Martin Brodeur's career win total is more impressive than Patrick Roy's, once you take into account the changes in league tiebreaking procedures (e.g. 4-on-4 OT, shootouts), the number of games played in a season, and the cyclical nature of how many games goalies play in a season. I decided to take a closer look at this question.

First of all, wins is not a very informative stat, unless you take into account the team a goalie is playing on. Goal support and the number and type of shots against all have a big impact on winning and losing. I will not be adjusting for any of those things in this post, so take the numbers with the standard critical eye reserved for goalie win totals. The objective here is to compare wins across different eras rather than to evaluate goaltenders.

First of all, for each season since 1917-18 I figured out the percentage of games that ended up in a win for one of the goalies, and used it as an adjustment factor in my analysis. In today's NHL, that figure is 100% because of the shootout. In 2003-04, before the shootout was introduced, 14% of games ended up as ties, so just 86% of the time a goalie was awarded with a win. In lower scoring eras early in the NHL's history, it was not uncommon for the percentage of games with a win to be at 80% or lower.

Another consideration was the number of games in a season. Everything was normalized to the current 82 game schedule.

The final thing I looked at was how goalies were utilized. I wanted to use a similar measurement for all the years, so I took the average of the top 6 goalies with the highest minutes played in each season. I then figured out the percentage of minutes played by the top 6, and used this as an adjustment factor. In seasons where the starters played nearly every game, such as the 1930s, the number was very close to 100%. Last season the figure was 89%. For much of the 1970s and 1980s, however, this number was much lower. For example, in 1986-87 the top 6 goalies played just 67% of the minutes.

The problem with the minutes adjustment was that it was quite variable. In the Original Six era, if one or two teams decided to platoon goaltenders for a season it would drastically reduce the overall total. I therefore decided to use a three-year average for each season, consisting of the previous season, the current season, and the following season. This smoothed the curve and allowed for, in my opinion, a better reflection of the era effect, rather than putting an excessive weighting on an individual team's roster management choices.

This allows us to calculate a season-by-season adjusted win figure, based on the win frequency, schedule length, and average level of minutes played for top goalies. Here are the all-time top 30 (up to the end of 2007-08):

RankGoalieWinsAdj Wins
1.Patrick Roy551610
2.Terry Sawchuk449558
3.Jacques Plante437555
4.Martin Brodeur538543
5.Glenn Hall407536
6.Tony Esposito423529
7.Ed Belfour484515
8.Curtis Joseph449474
9.Grant Fuhr403462
10.Rogie Vachon355447
11.Mike Vernon389443
12.Tiny Thompson284442
13.Gump Worsley335437
14.Andy Moog372434
15.Clint Benedict190430
16.Turk Broda302429
17.John Vanbiesbrouck374426
18.Harry Lumley330421
19.Tom Barrasso369415
20.Dominik Hasek389399
21.John Ross Roach219397
22.George Hainsworth246384
23.Ed Giacomin289381
24.Chris Osgood363373
25.Billy Smith305373
26.Frank Brimsek252364
27.Mike Liut294361
28.Sean Burke324352
29.Dan Bouchard286351
30.Bernie Parent271341

By this measure, Patrick Roy is still well ahead of everyone. Martin Brodeur not only has a lot more ground to make up on Roy, but he also has to first pass Plante and Sawchuk. Based on these figures and his current health situation, Brodeur probably won't pass Roy in career adjusted wins until some time in 2011.

These results show that wins were much easier to come by in the late 1990s and early 2000s. For example, coming into this season Roberto Luongo had a total of 197 wins but only 182 adjusted wins, which shows how win totals are being inflated in the current era where top goalies play nearly every game and somebody wins every single night. Of course Luongo is a textbook case of a goalie who is very underrated by wins totals because he has played most of his career on weak teams, but that is just a reminder of the limitations of the statistic.

Roy's wins record is certainly more impressive when you take into account the adjusting factors. There are many who make the mistake of comparing Brodeur's numbers directly with Roy's, but that is something that simply cannot be done without understanding the league contexts. Once you adjust for that, Patrick Roy comes out ahead in almost everything. There is no question that both Roy and Brodeur played most of their careers on outstanding teams, but at this point Roy still has to be rated the better goaltender, and that is even without taking into account their playoff records (where Roy has a large advantage over Brodeur).

This list also doesn't adjust for career length. This is a disadvantage for older goalies who tended to have shorter careers, both because they were more likely to get injured and also because there was a lot more competition for the few starting jobs that were available. That is likely the main reason that 7 of the top 10 goalies on the list played most of their careers post-1967. It is also why the longevity of Sawchuk, Plante and Hall is impressive, and why they are usually rated very highly on all-time goalie lists.


overpass said...

As you say, there still aren't many old-time goalies high on this list, but there would be at least one more if you included pre-NHL play. Clint Benedict played for 5 years in the NHA before the NHL was formed, and would certainly have over 500 Adj Wins with these years included.

Anyway, the list looks good, or at least as good as a list based only on wins can. I'd like to see Hasek's Adjusted Wins from the Czech league added to this, but that's well past the scope of this list. :-)

Anonymous said...

haha, haseks czech league numbers? whats next, counting guys pee wee hockey totals? this site is getting a little carried away with all these theoretical or "what if everybody played at the same time" totals. there are way too many variables to even somewhat accurately create "adjusted" totals. there is no way to measure how ken dryden would have done on todays detroit red wings, playing 76 of 82 games, using henrik lundqvist sized pads. sure we could use a bunch of stats to come up with a guess, but all of these "adjusted" stats fail miserably when tested using a basic statistics confidence interval.

Anonymous said...

man does this guy have a hard-on for discrediting brodeur. note how the calculation greatly adjusted wins for every goal but brodeur, giving him a +5 adjustment. cute.

Al said...

The biggest problem with this is that virtually none of these goaltenders couldn't have played nearly as many games per season as Brodeur does. He is one of the most durable goaltenders of all-time, and, as such, is really punished for that in these adjusted stats more than anything.

Anonymous said...

Sorry, that would be could not couldn't in my last post.

Anonymous said...

hey guys........i dont know what this author is smoking.....but marty now averages over 40 wins a season.........meaning it will not take him until 2011 to catch Roy's bs adj. wins number

Anonymous said...

and as of today brodeur has 642 wins... whoever wrote this article is a moron!!!!