Roy's prime seems to be generally considered to be 1988-1992. This can be justified by league-adjusted statistics and awards voting. However, with the team factors that likely helped Roy in Montreal, it remains questionable as to how much of a dropoff he actually experienced in his own play after 1992. There are some other factors that support this assertion:
A. League Average Save % Increased Near the End of Roy's Statistical Prime
Here's the league average around those seasons:
1990-91: .886
1991-92: .888
1992-93: .885
1993-94: .895
1994-95: .901
1995-96: .898
1996-97: .905
League average climbs rapidly starting in 1993-94. This was likely because of several factors, including better goaltenders entering the league, an overall switch to the more efficient butterfly style from standup,
better and bigger goalie equipment, increased competition for the playoffs with league expansion and the 1-8 conference format, and a generally more defensive style of play. Any metrics relative to league average implicitly assume that league average is directly comparable from one season to another. However, any outside factors that had more of an impact on the average in some seasons than others could have a skewing effect on adjusted scores.
B. The Best Goalies Did Not Make the Largest Save % Gains
It is informative to look at the results of goalies who were starters both before and after the sudden rise in league average save percentages. I used the NHL lockout in 1994 as the break point and looked at the combined periods of 1990-91 through 1993-94 and 1994-95 through 1997-98. Any goalie who played at least 150 games in both samples qualified. Here is the list:
Goalie|Birth Year|GP 91-94|SV% 91-94|GP 95-98|SV% 95-98
Ed Belfour|1965|267|.905|199|.906
Bill Ranford|1966|265|.890|194|.887
Patrick Roy|1965|245|.908|231|.914
Mike Vernon|1963|229|.885|157|.897
Curtis Joseph|1967|229|.909|213|.902
Andy Moog|1960|223|.888|162|.908
Kirk McLean|1966|212|.889|173|.888
Ron Hextall|1964|200|.891|185|.904
Mike Richter|1966|192|.901|209|.908
John Vanbiesbrouck|1963|190|.908|211|.908
Grant Fuhr|1962|169|.887|227|.899
Average|1964|220|.896|196|.902
League average from 1990-91 to 1993-94 was .889, while the average from 1994-95 to 1997-98 was .903.
While the league average jumped .014, the average of the goalies playing the most games only went up .006. Some of this would be age-related decline, and there are team factors involved for some of these goalies. But it is interesting that the goalies with the biggest improvement were actually the ones with the weaker numbers from the prior period. The two standup goalies who never really changed their style did not see their numbers rise much (Ranford, McLean). Some of the '80s goalies who incorporated more of the butterfly game and/or moved to more defensive teams as the '90s went on saw their numbers rise quite a lot (Fuhr, Hextall, Moog, Vernon). And the goalies who were the best to begin with, the ones who would continue to be rated among the best in the league throughout the decade did not see that much of an improvement in their numbers from the early '90s to the late '90s (Belfour, Joseph, Vanbiesbrouck, and even Roy who had an average increase). If it was all because of external factors that affected everyone equally (like goalie equipment or overall scoring environment), then we'd expect to see the top guys benefit just as much as the bottom guys.
Does that mean that Roy, Belfour and company were better goalies early on? Possibly. It's also possible that they lost a few of their relative team advantages while the rest of the league was adopting their superior styles of play and catching up to their performance level.
A lot of older goalies didn't survive the early '90s. Many were pushed out by all the young Quebecois goalie prospects entering the league, a wave of talent influenced by Patrick Roy. There was also a greater influx of Europeans following in the footsteps of Pelle Lindbergh and Dominik Hasek, which created an even tighter talent pool (although it should be noted that while the top group of goaltending was quite strong the overall talent depth still isn't comparable to today, since even in the mid-'90s there was usually a significant gap between the starter and backup on teams around the league). There was also better overall goalie coaching, as teams saw the results of the Francois Allaire-Patrick Roy partnership.
C. Patrick Roy Was Rated Highly and Played Well Throughout His Career
Subjective rankings of Roy do not seem to match up with him having a great early peak and then falling off to a much lower level. It is not unusual to find references to Patrick Roy as the best goalie in the world anywhere between 1993 and 2002. My recollection is that many people rated Roy as the best in the game ahead of Hasek all along, even though Hasek was the one winning Vezinas. I think they were wrong, but more wrong because they were underrating Hasek than overrating Roy. Roy was also usually the highest-paid goalie in the league, occasionally second to Hasek.
It is also likely that most people considered Grant Fuhr to be better than Roy until at least 1990. See the 1987 Canada Cup, the 1987-88 All-Star and Vezina voting despite a large save percentage gap in Roy's favour, and
this player poll taken in January 1990 that gives a 40-17 edge for Fuhr over Roy. Granted, this is likely in large part a case of people focusing a little too much on championships, as they might have done during Roy's later career.
I rate Patrick Roy's 1993-94 regular season very highly. It was the second-best of his career in GVT, and he did it on a worse defensive team while facing more opposing power plays than on the Pat Burns Habs teams. It was good enough to win a Vezina in an ordinary year, it was just Roy's bad luck that Hasek and Vanbiesbrouck also played outstanding that season.
Considering that and his 1993 playoff run, I don't think it makes much sense to claim that Roy's prime ended before 1994. And it's hard to know what to make of him in 1995 and 1996. The lockout year was a shortened season on a weak team. Hasek in 1996 had a similar large save percentage drop during his prime also on a non-playoff squad. The following year had all the turmoil of Le Trade (Roy's 9 GA outing against the Red Wings took his seasonal save percentage down by .004 alone), and Roy did follow it with a great playoff run. By 1997 he was back up to a .923 on Colorado, a great season that again could have won a Vezina if he did it in another season (like 1995-96). From then on he continued to post generally strong numbers in Colorado. As Quoipourqoui pointed out, Roy's five year playoff save percentage peak was 1993-1998. Was Patrick Roy a worse goaltender in 1997 than he was in 1994 or 1992? I think it's at least possible that he wasn't, or at least not very far behind, but that the version of Roy playing behind the Montreal defence looked better relative to everyone else than he would later on.
Note that during Dominik Hasek's prime (1994-1999), Roy was the #2 goalie in the league in save percentage.
1. Dominik Hasek, .930
2. Patrick Roy, .915
3. Martin Brodeur, .914
League average over this period was .902, indicating that Roy was still well above average during this period, particularly since he had less of a shot quality advantage at this point of his career (and faced significantly more power plays against than Brodeur).
On first glance this result pales in comparison to Roy's .909 vs. .883 league average from 1988 to 1992. Yet assume that Roy got a +.007 for playing on Montreal relative to an average goalie (see my last post), and the difference between him and league average is .019, or .006 ahead of where he was from '94 to '99. Is it possible that the improved talent pool, widespread adoption of the butterfly, increased prevalence of defensive systems, better professional coaching and increased competition for playoff spots made an impact of around .006 or more on league average from 1994-99 relative to 1988-1992? I think it is possible. In any event, it seems unlikely that Roy's level of play dropped off by such a massive amount from the early '90s to the late '90s.
I'd say Roy is defined more by elite consistency rather than an all-time great peak. He was very good throughout his career with very few down seasons, all the way through 2002-03. Note also though that Dominik Hasek, outside of 1994-1999, had a .914 save percentage compared to a .903 league average, even though 40% of his sample came at an older age than Roy at his retirement, and nearly all of it was while competing against a deeper talent pool than the 1980s.
D. Overall Regular Season Career Value
Here are the current career GVT numbers in the regular season:
Patrick Roy: 1029 GP, 431.1 GVT
Dominik Hasek: 735 GP, 410.1 GVT
Also, Brodeur, just for reference:
Brodeur: 1191 GP, 307.4 GVT
If we split those up into prime and non-prime, defining Roy's prime as 1988 to 1994 and Hasek's as 1994 to 1999, here is how the two match up:
Patrick Roy, prime: 372 GP, 197.6 GVT, 0.53 GVT/GP
Patrick Roy, rest: 657 GP, 233.5 GVT, 0.36 GVT/GP
Dominik Hasek, prime: 361 GP, 278.9 GVT, 0.77 GVT/GP
Dominik Hasek, rest: 374 GP, 131.2 GVT, 0.35 GVT/GP
I think the overall indication of those numbers is probably about right, in that it shows Hasek's prime significantly surpassing Roy's prime, with the two of them roughly equivalent on a per-game basis in their non-prime performance and the main overall difference being that Roy played an extra 283 games.
However, it seems also clear that Roy's prime numbers are at least somewhat inflated. If we subtract even just 10% of his prime GVT because of shot quality or because of his likely benefits from using league average as a benchmark, that eliminates almost the entire career gap between Roy and Hasek in GVT.
That's without giving any extra credit to Hasek for his peak advantage, or factoring in a single game that he played internationally. Even if you think that Hasek would have been a starter but not a Vezina candidate in the NHL in the late '80s, that still would have meant that his overall career value would have easily surpassed Roy's by this measure. The evidence suggests that with Roy's Montreal years appropriately placed in context, Hasek had not only a significantly superior regular season peak but probably also an equivalent overall NHL regular season career.
Just got the PM that voting is open, hope to have a final post up by tonight on playoff performance relative to expectations.