Monday, July 17, 2006

One Method To Determine Best Goalies of All Time

I am returning to the topic of sabremetrics and hockey where I left off last summer. I had looked at two ways to rank the careers of various players. One was produced by pnep and one by Daryl Shilling. I had posted their comparisons of the top 10 forwards of all time and the top 10 defencemen of all time. I posted pnep's goalie ranking system but I never posted one from Daryl Shilling. In part, this was because Shilling produced a brand new hockey project ranking system that addressed forwards and defencemen. I hoped that a new goalie system might be coming along soon.

While there was no new goalie system, on his website, Shilling does have a a very good system (this system is an earlier version of Pnep's system. The credit goes to Pnep - Roman Nepomnyaschev).

Here is the way this system works:

Like the monitor for position players, the system generates an integer based on the goaler’s “standings” in certain key areas. The formula is easy to apply:

a) Award 50 points for being the starting goaltender on a Stanley Cup winning team
b) Award 25 points for being the starting goaltender on a Stanley Cup runner-up
c) Award 20 points for each selection to the All-Star game
d) Award 150 points for winning the Calder Trophy
e) Award 70 points if the goaltender was runner-up for the Calder
f) Award 80 points for being named to the 1st All-Star Team
g) Award 50 points for being named to the 2nd All-Star Team
h) Award 70 points for winning the Vezina Trophy
i) Award 55 points for being the Vezina Trophy runner-up
j) Award 30 points for winning the Jennings Trophy
k) Award 20 points for being the Jennings Trophy runner-up
l) Award 150 points for winning the Conn Smythe Trophy
m) Award 300 points for winning the Lester B. Pearson Trophy
n) Award 250 points for winning the Hart Trophy
o) Award 150 points being the Hart Trophy runner-up
p) Calculate: (Wins/2 + (Wins – Losses) + (4 * Playoff Wins))


This system overemphasizes what others thought of the goalie (ie. did he win awards, make all star teams?) as opposed to what he did on the ice. This may be inevitable, because goaltending statistics are hard to analyze. As I wrote discussing pnep's system:

Goaltenders are the hardest of the three positions to evaluate statistically. Most goalie stats are strongly biased by the team the goalie plays on. Good teams win lots of games, yet wins are considered a key goalie stat. Teams that allow few shots and few good scoring chances tend to have goalies with lots of shutouts and low goals against averages. These stats which are highly team dependant are used to evaluate goalies. Even saves percentage is biased by the team the goalie plays on. If a team allows lots of low quality shots, their goalie will tend to have a high saves percentage. Further, GAA's and saves percentages of goalies have changed dramatically in different eras of the NHL. Its not easy to analyze goalie stats in any meaningful way, but nevertheless here is pnep's attempt.

This system does incorporate the goaltending statistics to some level. In includes wins and losses and playoff wins. These are largely team statistics. While a goalie can go a long way toward a team winning, without a good team behind him, a good goalie (ie Roberto Luongo in Florida) will not get a lot of wins. The William Jennings points (and early year Vezina points) were statistical also. They were given to the goalie on the team with the best goals against average. A strong team defence will help a goalie win these awards. A Roberto Luongo type will never have a chance at them.

The rest of the numbers are what others thought of the goalie in question. Did they put him on all star teams, give him awards etc? This assumes that the people who select these awards are correct. Sometimes they make good choices, but other times they make mistakes. Those mistakes will carry into this system.

Also, this point system is cumulative. The longer a goalie plays, the more points he will get. This will serve to select goalies with long careers instead of those with shorter careers. While it is true that elite goalies tend to be able to play in the NHL well beyond their prime, sometimes they do not (as in the cases of Ken Dryden or Bill Durnan for example). Should this goalie be ranked below somebody who was not as good in his prime, but hung on longer?

I will look at the results of this system in the near future. Which goalies does it rate the highest?

NOTE: This is an early version of Pnep's Hall of Fame monitor for goalies. So it produces nearly the same results as his more modern ones.

Comments:
I've long felt that we're going to have to go back and calculate SP for the years 27-90 (or so) because:

1. It is a vital G stat
2. The line has moved back and forth from .900 over the years.

I've done some (very little) work in the late 60s and was surprised to find 30 games stretches of .900+ while the mid-80s saw the bar drop 20 points below that (to my eye).

Still, a ratings system that puts a premium on leading in SP for a specific year would be more accurate imo.
 
Saves percentage might be more accurate, but it wouldn't be perfect. Some teams will allow higher quality shots than others and this is not taken into account.
 
Lowetide, check out "The Hockey Compendium" by Jeff Z Klein and Karl-Eric Reif. One of their readers calculated save percentage for each goalie from 1955-1967 and a few years in the 70's. The complete results are published, in addition to 1983-present. It's an amazing read.
 
Hey !!
its my research (early version "Goalie HHOF monitor") :)

Darryl quote:" I present the results here, but the work here is completely and totally Roman’s baby."
 
Pnep

I am sorry. I didn't read all the fine print properly. This is an earlier version of your goalie HOF monitor. Sorry for not attributing credit properly. I'll fix it.
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?