Age and Goaltender Regression
- Thread starter StefanW
- Start date
Hi, read your link and liked it but had some issues with it...
1) Recently also did a study on goalies and there peak years and it roughly agreed with conclusions of the study you quoted. I only researched goalies ages 20-30 for the last decade but I didn't base my study on save percentage which while the sample size was much smaller adds some extra weight the the study you quoted.
2) Your study strongly implies that the team with the best goalie will win the cup or at least win several play-off rounds. I'm not sure this is a great assumption although I really like the idea of weighting Play-off data heavily and the fact you aren't using save percentage. I'm of the opinion the improved quality of competition offsets the small sample size more than most people.
3) I posted a couple weeks/months ago that given it is difficult to spot good goalies it makes sense that teams stick with them longer than there primes due to good track records instead of replacing with what is effectively a wild card.
At the very least I would look into correcting for number of goalies of a given age in the league at the time. I suspect your numbers are being skewed because of this.
4) If I had to judge goalies on one stat that is commonly collected I can understand the argument for Save Percentage and would even admit that in a big enough sample size all it's weakness are averaged out in the wash. However I would argue that given a subject where very little is know or understood I try and measure something closer to what I'm actually want to measure. All things being equal what you want is a goalie who wins the most games failing that you'd want a goalie who lets in the fewest goals failing that Save Percentage would probably be next.
Save Percentage is a first attempt adjust for the fact all things are not equal but I would say that it isn't actually what you are trying to figure out.
p.s. Not surprising best Save Percentage didn't win Cup often, to a certain extent I'd say Save % records how heavily a team is leaning on there goalie for success and opposite for deep teams.
1) Recently also did a study on goalies and there peak years and it roughly agreed with conclusions of the study you quoted. I only researched goalies ages 20-30 for the last decade but I didn't base my study on save percentage which while the sample size was much smaller adds some extra weight the the study you quoted.
2) Your study strongly implies that the team with the best goalie will win the cup or at least win several play-off rounds. I'm not sure this is a great assumption although I really like the idea of weighting Play-off data heavily and the fact you aren't using save percentage. I'm of the opinion the improved quality of competition offsets the small sample size more than most people.
3) I posted a couple weeks/months ago that given it is difficult to spot good goalies it makes sense that teams stick with them longer than there primes due to good track records instead of replacing with what is effectively a wild card.
At the very least I would look into correcting for number of goalies of a given age in the league at the time. I suspect your numbers are being skewed because of this.
4) If I had to judge goalies on one stat that is commonly collected I can understand the argument for Save Percentage and would even admit that in a big enough sample size all it's weakness are averaged out in the wash. However I would argue that given a subject where very little is know or understood I try and measure something closer to what I'm actually want to measure. All things being equal what you want is a goalie who wins the most games failing that you'd want a goalie who lets in the fewest goals failing that Save Percentage would probably be next.
Save Percentage is a first attempt adjust for the fact all things are not equal but I would say that it isn't actually what you are trying to figure out.
p.s. Not surprising best Save Percentage didn't win Cup often, to a certain extent I'd say Save % records how heavily a team is leaning on there goalie for success and opposite for deep teams.
Last edited:
StefanW
Registered User
First off, thank you very much for taking the time to read and respond. Feedback is always very valuable. I'll try to address the four listed points/criticisms to the best of my ability.
Yeah, the study I quoted was rock solid, so I assume the association between age and save percentage that was laid out is completely accurate. If you got the same result you are definitely doing something right.
I believe starting assumptions, rooted in some sort of theory, are the absolute key to analyses such as these. My theory that a goalie has to be excellent to lead his team into the final four may be flawed to be sure, but I believe it is less flawed than an aggregate yearly save percentage. Reasonable people can disagree on points like this, and such disagreements are great starting points for discussion.
I mention larger possible trends in goaltender ages a couple of times, but you are correct in pointing out that I do not correct for this. The problem is my sample is 120, and when you divide by the three eras you end up with 40 per decade. When the subsamples are so small any type of correction would add loads of error, so I avoid doing it. My preference is to point out that younger goalies were in vogue 20-30 years ago and let the reader tick that off as a point of interest or possible flaw. This flaw stems from a conscious decision made after weighing my options, so mea culpa.
My points were that: 1) save percentage are commonly used in the analytics community as a proxy for "good goalie", which is why goaltender regression is measured by drops save percentage over time; 2) "good goalie" can be measured in many alternate ways, including how far a goalie takes his team into the playoffs; and 3) the difference between the two positions is theoretical (i.e. both can do good math and come up with entirely different conclusions).
I see the article I wrote as one point on the road to creating better analytics for goaltenders. Currently, hockey analytics are very offensive minded. However, defense can win games and championships as effectively, and sometimes more effectively, than good offense. Goaltending analytics need to be developed to the point where they are at least up to speed with possession measures on offense and things of that nature. It is currently a gaping hole in hockey analytics.
I chalk it up to the difference between being steady all year, which gets you into the playoffs and helps your seeding, to getting hot in the playoffs. If in-their-prime Marty Brodeur or Patrick Roy got hot in the playoffs they had a excellent shot of winning the cup regardless of how the regular season went. It takes 8 wins to make the final four, and 16 wins to take home the cup. That is very different from a long season, and hitting a hot streak at the right time matters.
Yeah, the study I quoted was rock solid, so I assume the association between age and save percentage that was laid out is completely accurate. If you got the same result you are definitely doing something right.
I believe starting assumptions, rooted in some sort of theory, are the absolute key to analyses such as these. My theory that a goalie has to be excellent to lead his team into the final four may be flawed to be sure, but I believe it is less flawed than an aggregate yearly save percentage. Reasonable people can disagree on points like this, and such disagreements are great starting points for discussion.
I mention larger possible trends in goaltender ages a couple of times, but you are correct in pointing out that I do not correct for this. The problem is my sample is 120, and when you divide by the three eras you end up with 40 per decade. When the subsamples are so small any type of correction would add loads of error, so I avoid doing it. My preference is to point out that younger goalies were in vogue 20-30 years ago and let the reader tick that off as a point of interest or possible flaw. This flaw stems from a conscious decision made after weighing my options, so mea culpa.
My points were that: 1) save percentage are commonly used in the analytics community as a proxy for "good goalie", which is why goaltender regression is measured by drops save percentage over time; 2) "good goalie" can be measured in many alternate ways, including how far a goalie takes his team into the playoffs; and 3) the difference between the two positions is theoretical (i.e. both can do good math and come up with entirely different conclusions).
I see the article I wrote as one point on the road to creating better analytics for goaltenders. Currently, hockey analytics are very offensive minded. However, defense can win games and championships as effectively, and sometimes more effectively, than good offense. Goaltending analytics need to be developed to the point where they are at least up to speed with possession measures on offense and things of that nature. It is currently a gaping hole in hockey analytics.
I chalk it up to the difference between being steady all year, which gets you into the playoffs and helps your seeding, to getting hot in the playoffs. If in-their-prime Marty Brodeur or Patrick Roy got hot in the playoffs they had a excellent shot of winning the cup regardless of how the regular season went. It takes 8 wins to make the final four, and 16 wins to take home the cup. That is very different from a long season, and hitting a hot streak at the right time matters.
Agreed with the above. Posting messages always dicey was trying to give you credit for the above.
Sorry about that read previous day and posted today. Glancing over it again I was curious about a break down of ages of goalies in league or play-offs or... Though it might be something that you could quickly gather or comment on and/or speculate on how much this affects your other graphs. I always enjoy taking the next step from raw data to wild speculation.
I don't disagree with anything you said here but if you think about it I'm pretty sure the affects that your taking about above would match up with the Sv% comments I made. You've given causes and I listed the affect.
Bad teams would need goalie to be consistently on top of his game and couldn't let in many or meaningless goals. Also bad teams would probably allow more shots (believe there's an old study of more shots correlates with higher save percentage). Etc, either we agree that we agree or don't at this point.
