(Hopefully this post is not too late to be discussed prior to this round of voting. If people think that the type of numbers I'm producing here are worthwhile I can continue running them for goalies in subsequent rounds as well.)
Patrick Roy has the highest level of overall achievement in the playoffs, according to traditional metrics like wins as well as more advanced stats like GVT. That is to his credit, but at the same time it shouldn't be that surprising as Roy played in the era of four playoff rounds, he played on consistently strong teams, and he was one of the best goalies of all-time.
A vitally important factor to consider for any goalie's playoff record is opportunity. There is a reason that Ken Dryden's 1971 Stanley Cup is celebrated far more than any of his Cups from 1976 to 1979. Taking a team farther than expected in the playoffs is what builds reputations. It is in fact one of the main arguments that people use to claim Roy is the best goalie of all-time (referring mainly to 1986 and 1993).
I do not generally put a lot of emphasis on team success. However, my philosophy is that if you are going to award points for team success, at least make sure that you account for the goalie's opportunity and expectations, and also make sure that you look at their complete record. It is not enough to know what kind of teams a goalie had when they won, but also what kind of teams they had when they lost. Opposing matchups are also important, as high seeds can cruise through early rounds while low seeds usually need to repeatedly upset top teams to make the Finals.
The Method
I developed a playoff prediction model, using
Bill James' Log5 formula, which assesses the likelihood of a team winning a head-to-head matchup given their winning percentage and the record of their opponent. The formula (expressed in a slightly simpler way than in the link above) is: (W% Team A - W% Team A * W% Team B) / (W% Team A + W% Team B - 2 * W% Team A * W% Team B). This formula gives the probability of Team A winning any given game. That probability can then be used to assess the likelihood of either team winning a seven game series.
This assumes that the probability of winning each game is identical, which is not technically a correct assumption as it does not account for home-ice advantage. This therefore likely slightly underestimates the probability of the favoured team winning.
Example Calculation
I'll use Ken Dryden's 1971 playoff run as an example. The 1971 Canadiens had a .622 winning percentage, while the 1971 Bruins were at .776. The formula gives Montreal's expected win percentage in a one game matchup as .322. The probability of a team with a .322 win probability in each game winning a seven game series is just .156. That Dryden's Habs won that series represents a very significant upset, as everyone is well aware. The next round against Minnesota (.462) was much more routine, with an .812 expected series win probability. The Cup Final against Chicago (.686) was again an upset, as the Habs had a 34.9% chance of winning.
From this we can figure out the probability of the team winning any given number of playoff rounds.
Probability of 1 series win: 0.156 * (1 - .812) = 0.029
Probability of 2 series wins: 0.156 * 0.812 * (1 - .349) = 0.082
Probability of 3 series wins: 0.156 * 0.812 * 0.349 = 0.044
Expected playoff series wins: (.029) + 2(.082) + 3(.044) = 0.33
In 1971, Dryden's team won 3 playoff series when they were expected to win 0.3, which is outstanding. Of course that was just one season, in 1972 the Habs lost in the first round in a series that was roughly 50/50 (.482), and as a result won zero rounds with an expected number of 0.81. The following year the Habs again won the Cup, but given that they were heavy favourites in each round this time the team had a 71% chance of winning it all and Dryden's expected number of series won was a whopping 2.4. Anything less than a Cup victory would have been a failure for the 1973 Canadiens, and therefore Dryden gets much less additional credit for that victory.
The final step is to run the numbers for every season of the goalie's playoff career, to get their expected number of playoff series wins and Cups, and then compare them to the goalie's actual achievements.
These numbers can be easily calculated for any era of the NHL. However, the reason this is possible is that it is not actually dependent on any goaltending statistics. Dryden's goaltending did happen to be a major factor in the Canadiens' 1971 Cup win, but if his play had been average and the Canadiens won because they suddenly started scoring 6 goals per game then he would have still gotten the same credit. It is useful to compare the expected results with a goalie's statistical playoff performance to see if they are in line, or if perhaps the rest of the team might be more responsible.
The other caveat is that it assumes regular season performance is a good predictor of playoff success. Note that this is generally true. The team finishing first overall is the most likely to win the Cup, then the team finishing second, etc. Very few low-seeded teams have ever won a Cup, the 2012 Kings being a notable exception. The main problem with using regular season results is that the playoff version of the team is not always the same as the regular season version. Coaching changes and trades or other roster moves can change a team's fortunes in mid-season, and make them more formidable than their seeding suggests come playoff time. For recent years, shootout results are included in regular season numbers, which adds an element of luck to a team's standing results and probably results in poorer predictions.
One other important thing to note is that a goalie's own performance is included in the team's regular season record. Therefore, a goalie that elevates a weak team or a goalie that plays poorly and makes a good team appear worse than they are is in a sense biasing their own results, or at the very least being judged against their own usual standard. I will come back to this point later.
The Numbers
Here are the goalies currently under consideration, ranked by the total number of playoff series won above expectation:
Rank|Goalie|Series W|Exp Series|Diff|%|Cups|Exp Cups|Diff|%
1|Patrick Roy|32|23.95|+8.05|+34%|4|2.07|+1.93|+93%
2|Ken Dryden|19|14.27|+4.73|+33%|6|3.42|+2.58|+75%
3|Dominik Hasek|12|8.48|+3.52|+42%|1|0.67|+0.33|+49%
4|Martin Brodeur|23|20.87|+2.13|+10%|3|1.53|+1.47|+96%
5|Terry Sawchuk|11|9.97|+1.03|+10%|4|3.36|+0.64|+19%
6|Jacques Plante|14|14.56|-0.56|-4%|5|5.43|-0.43|-8%
7|Glenn Hall|7|8.38|-1.38|-16%|1|2.56|-1.56|-61%
A few notes about what was included:
Any time a goalie near the very end of his career in his very late thirties or forties did not play much in the playoffs, I excluded it from his numbers (Sawchuk in '65, '66 and '70, Hall 69-71, Plante '70-73, Hasek '08). Similarly, I excluded it whenever a goalie had not yet established himself as the starter and did not play the majority of the playoff games (Plante '53, Hasek '91-93, Brodeur '92). I also did not include any season where a goalie did not play at all because of injury, although Hasek's numbers are included for 1997.
There may also occasionally have been some ambiguity in defining which goalie got credit for the playoff series win, in cases where more than one goalie participated in the same series. It was usually determined to be the goalie who played the most in that series, although if there was a goalie who played most of the minutes during the entire playoff and shared multiple starts in a single series I would generally include that series as well in their numbers.
Analysis:
- Patrick Roy ranks first with the highest difference between actual series wins and expected series wins. However, he also ranks first in expected series wins, with nobody other than Brodeur even close. In percentage of actual vs. expected Roy was virtually tied with Ken Dryden, and Roy ranks second in both most total Cups won above expected and percentage of Cups won above expected. Roy's playoff team success results are very good, and his total numbers are the best in the group. However, he doesn't stand nearly so far ahead of the pack on a percentage basis, because he should have won a lot of playoff series given his long career and the teams he played on.
- Dominik Hasek finished first in percentage of playoff series won compared to expected. Hasek also had by far the lowest score in expected Stanley Cups with just 0.67, while every other goalie was expected to win at least 1.5.
- Martin Brodeur was perhaps surprisingly the goalie with the highest percentage of Cups above expected, narrowly edging Roy, although there was a clear gap between Brodeur and the leaders in terms of playoff series won. Brodeur's Devils had a tendency to either lose early or go all the way. The '95 Devils were also probably underrated by their regular season results over the shortened lockout season, given the way they dominated their opponents in the playoffs.
- Terry Sawchuk ranks even with Brodeur in percentage terms for playoff series wins, and comes in slightly above expected for Cups won. This supports the general perception that he has a fairly strong playoff record.
- Jacques Plante's results are surprisingly underwhelming. Outside of the Canadiens' five Cups in a row in the 1950s his teams did very little in the playoffs. Plante also never won a single playoff series in which his team did not have a better regular season record than their opponent, although there were only four of those as his team was the favorite nearly all of the time.
- Glenn Hall's team success numbers were easily the poorest of any goalie in this group. How much Hall himself was to blame has yet to be determined, but Chicago repeatedly came up short in the playoffs in the 1960s.
- This, in short, is the basic argument why Roy's playoff career numbers are not the absolute difference-maker that requires him to be ranked ahead of Hasek. In terms of save stats they two were very close on a per-game basis (.01 apart in GVT/game, Roy with a slight edge in adjusted save percentage but that's not accounting for his Montreal defence or the lower league average in the late '80s), and if their relative team success is also similar, then really all we're doing is crediting Roy for extra opportunity, and that is not enough to offset Hasek's regular season prime advantage.
- There are two additional issues in the Roy vs. Hasek playoff debate that have been brought up as well: The injury factor with Hasek, and the impact of regular season success on a team's winning record, particularly when a goalie himself may have made the difference between making or not making the playoffs in the first place. I've got a few more numbers coming on these two topics as well.