So, the reason I asked you this in the first place, was to respond to this sentence:
Meh, maybe...I mean, let's say it's not tied directly to volume somehow...are we going to pretend like it's tied to talent all of a sudden (well, estupidos already do this) and not account for team effects and the like?
Until all appeals have been filed with the appellate and supreme courts, I believe the first part of this sentence is sufficiently addressed. There is no actual connection between shot volume and save percentage, either game-by-game (0.08), or on the aggregate (-.02) (and, the correlation between save percentage and win percentage is actually fairly high, as you may recall: 0.60). But you've submitted on that, far as I can tell.
When you said save percentage was not tied to talent, I wanted you to prove it. Because after all, save percentage is tied to winning, and goalie talent has to be tied to winning, so it's likely that save percentage and goalie talent are tied to eachother to some degree as well, in theory.
I wanted to know how
you would rank the most-used goalies of the past 20 years, and then compare your list to their prime save percentages, compared to league average. Here's what I did:
I took those 50 goalies and reviewed their careers individually. In each case I viewed them at their best. 7 consecutive seasons. Whichever 7 seasons would yield the most favorable result to them, based on hockey-reference's GA%-, which essentially is just how many points away from the league average save percentage they were.
Then I took your list of 50 goalies based on tiers, and gave them skill rankings in each tier: 100, 97, 94, 91, 88, 85, 82, and 79. These numbers are completely arbitrary. They seemed logical to me, being that if the best in the league is a guy who's a 100, then replacement level must be somewhere around 75. In any case, I don't think it matters all that much because the question is, does a higher number in the prime save percentage column tend to mean a higher number in the Mike Farkas Skill Assessment column?
I believe that the answer is yes. The correlation that I determined was 0.49. It's not negative, it's not weak, on the other hand it cannot be described as "direct" or "powerful". I would lean towards "fairly strong" as a good descriptor of what this means. I'd say that jives pretty well with what I've found here:
How to Interpret a Correlation Coefficient r - dummies
I think that it should also be noted, that if you put Tim Thomas in a skill range closer to where his numbers would indicate, such as in tier 2, the coefficient of correlation becomes 0.57. that's right, in a data set with 50 points, the fact that you put the guy with the third best numbers in the worst tier was enough to tip the correlation that much.
Here's a picture of the scatterplot. The fact that you have the goalies in tiers and not ranked one by one means that many of them end up in the same place vertically, which gives this an odd look... but the upward slope is visible for sure. although there are minor exceptions, look at the emptiness in the top-left and bottom right. You don't think anyone with crappy sv% is very good, and you don't think anyone with a great sv% is bad.... except for Tim Thomas, that is.
The other thing worth noting here is, all of these guys got into a lot of NHL games - that's exactly what the list was based on. Top-50 in the last 20 years. These were all pretty decent goalies, with only a few exceptions. Only six failed to post a save percentage above the league average over their best 7 consecutive seasons. If I asked you to rank the top-50 of the last 5 years, 20 of those would be below-league-average statistical goalies. The worst statistically in that time are Hiller, Scrivens, Markstrom, Ward, Niemi and Chad Johnston. I already know half those guys you don't like, and I have a feeling you're not fond of the other half. Meanwhile, on the other end... Price, Rask Bobrovsky, Luongo...
So, I believe that in a sample that doesn't have the selection bias that I introduced (only include guys that coaches saw fit to put into 366+ NHL games), the correlation between their numbers and your skill rating would be much, much higher. I don't know if I feel like testing that, plus now the purpose and methodology of the study has been explicity explained to you, which introduces the potential for your own bias, so I don't know if it would be that useful to all of us.
In any case, I think that what this shows is that you're not giving save percentage due credit for what it says about goalies. It agrees with you much more often than it doesn't.
And lastly, keep in mind that this is based just on
raw save percentage, the best "simple" goaltending statistic that you'll find on the back of a hockey card. But, many people smarter than either of us have done great things with save percentage such as: shot quality adjustment, special teams adjustment, save percentage when tied or winning by one, and so on. I believe that for the most part, analyses like this only improve the stat from its raw form, and would bring the two of you even closer together than you already are. I do not think it looks good on you to be as dismissive of save percentage as you are.