What sv% would a goalie of peak Hasek's caliber obtain in today's NHL?

[Moops]

During Hasek's six-year statistical prime (1993-94 thru 1998-99), he complied a .9296 sv% over a sample of nearly eleven thousand shots on goal. Those six seasons alone, he complied 296.5 goals saved above average, or enough to add something like 18-19 points more than an average goalie would add in the standings.

He was .027 better than the league average save percentage. In today's NHL, 27 goals above average/1,000 shots would come out to a sv% of .940--this would be not only impressive, but borderline impossible. There are plenty of unstoppable pucks that go into the net through no fault of the goaltender. I doubt that even the modern day version of Hasek could ever sustain a .940 over 11,000 shots.

(I say modern-day Hasek, because today's "perfect" goaltender would have a drastically different style from the real Hasek.)

On the other hand, I am sure that Modern Hasek would be worth a lot more than a .930 peak/.922 lifetime save percentage. If we had a generational supertalented goaltender coming into play this year, how good could this goaltender be?

I suppose the larger question is, are we past the days where world-class goalies can reach that level of all-time transcendence? We have so many good goaltenders that it's becoming almost impossible for a goaltender to be great. My best guess is that an ideal goaltender could realistically hope to peak at about 20 goals/1,000 above average.

Click to read more...

 

[Doctor No]

I can ballpark something...

From an outlier perspective, Hasek's best season was his 1993-94 campaign, where he stopped 93.0% of 1552 shots. This was 4.6 standard deviations above the league average (excluding Hasek).

What would a 4.6 standard deviation performance look like in 2013-14, where the league average save percentage was 91.38%? Facing 1552 shots, Hasek would stop 1469, or 94.7% of shots faced.

Of course, this is based on his one-year best, not his six-year span.

 

[TheDevilMadeMe]

I can ballpark something...

From an outlier perspective, Hasek's best season was his 1993-94 campaign, where he stopped 93.0% of 1552 shots. This was 4.6 standard deviations above the league average (excluding Hasek).

What would a 4.6 standard deviation performance look like in 2013-14, where the league average save percentage was 91.38%? Facing 1552 shots, Hasek would stop 1469, or 94.7% of shots faced.

Of course, this is based on his one-year best, not his six-year span.

I think the OP's point is that with goaltending, there is a hard limit as to what a "perfect" performance would be - 100%. Compare to a scoring forward - it's always theoretically possible to score more points if a better player comes along.

So as average save percentages get closer to the limit, it would make sense that it might become more difficult to be a certain number of standard deviations above, correct? Or are you calculating SDs based on error rate in a way that takes such a concern into account?

 

[WhalerTurnedBruin55]

I know shorter sample size but... .938

Tim Thomas did it in in 2010-11.

I'd say in order for a modern goalie to come up with numbers to be akin to Hasek's peak, it would basically need to match Tim Thomas's numbers over more seasons and more games per season.

 

[wintersej]

Correct me if I am wrong, but wasn't even strength scoring in the 90s pretty similar to even strength scoring now? They just had a crap load more power plays. I would think correcting for the extra powerplay time (and its lower save percentage) would get you pretty close.

 

[Cool Bryz]

It's an EXCELLENT question. You would probably need to estimate how many "unstoppable" shots there are over a large sample size (ie by looking at film). Somehow I don't think this fraction is as high as 6%.

 

[hatterson]

Correct me if I am wrong, but wasn't even strength scoring in the 90s pretty similar to even strength scoring now? They just had a crap load more power plays. I would think correcting for the extra powerplay time (and its lower save percentage) would get you pretty close.

According to the data I have Even strength scoring in 93-94 was at .049 goals per even strength minute and games contained roughly 89.43 even strength minutes.

In 11-12 (the last 82 game season I have the data for, I'll add in 13-14 soon) even strength scoring was at .041 goals per minute and games contained roughly 99.27 even strength minutes

So ES scoring is indeed lower (along with also having fewer PP minutes) although I haven't looked at save percentages versus shot rates to see why scoring is lower.


https://docs.google.com/spreadsheets/d/1yI6CYGsA1YSxlrcZaXdaS5TecZoufblpYfHtWOXn_hY/edit#gid=0

 

[Doctor No]

I think the OP's point is that with goaltending, there is a hard limit as to what a "perfect" performance would be - 100%. Compare to a scoring forward - it's always theoretically possible to score more points if a better player comes along.

So as average save percentages get closer to the limit, it would make sense that it might become more difficult to be a certain number of standard deviations above, correct? Or are you calculating SDs based on error rate in a way that takes such a concern into account?

True - and the binomial distribution's standard deviations aren't letter-perfect as we approach the boundary conditions.

I'll run some simulations, and see if there's something to it in this situation.

Sniff-wise, 94.7% sounds fair to me (admitting that it was a ballpark estimate).

 

[Doctor No]

Here's my simulation exercise. I'll also post my (very simple) R code, in case someone wants to play along. Don't enter the parentheticals.

What were the conditions in 1993-94 (other than Hasek)?

n = 1552
(shots faced by Hasek in 1993-94)
p = 1-((6927-109)/(66017-1552))
(league-wide save percentage in 1993-94, excluding Hasek)
h=rbinom(100000000,n,p)
(creates 100 million sample seasons, facing 1552 shots, with a league-average save percentage)
table(h)
(produces a summary distribution of the sample)

In my simulation, of the 100 million seasons, how many times did the goaltender stop 1443 of 1552 shots (or more)? 113 times, or 0.000113% of the time.

(As an aside, the "best" season I saw had 1451 saves, or a save percentage of 93.5%. The "worst" season I saw had 1315 saves, or a save percentage of 84.7%. Your results will vary).

Now, what's the corresponding cumulative distribution in the 2013-14 season?

n = 1552
(shots faced by Hasek in 1993-94)
p = 1-(6349/73683)
(league-wide save percentage in 2013-14)
h=rbinom(100000000,n,p)
(creates 100 million sample seasons, facing 1552 shots, with a league-average save percentage)
table(h)
(produces a summary distribution of the sample)

In this 100 million season sample, counting from the best seasons downward, how far do I have to go to get 113 seasons?

1467 saves, or a save percentage of 94.5%.

My conclusion is that Hasek's 93.0% save percentage in 1993-94 translates to approximately a 94.5% save percentage in 2013-14.

Note that this is based solely on the results of one (large) simulation exercise, and I could have gotten an "exact" answer using the pure distribution functions - but I prefer simulations. This also has a few simplifications that have been discussed in this thread (would Hasek have faced more or less power plays, more or less "unstoppable" shots, et cetera). Your mileage may vary.