Hockey Analysis Discussion

[butterfly]

This will likely be boring for most people but I thought we should have a thread to discuss hockey analysis generally.

I just subscribed to Evolving Hockey ($10/month for everything including data exports) and I kind of feel like a kid in a candy shop.

It's interesting that almost everyone is using the NHL's RTSS data so analysis is limited by what they give us. I'm sure they know about pre-shot movement and positioning but it isn't available.

They are using xGA for the defensive component in their model to attempt to isolate the effect of goaltending and the GF RAPM for the offensive component, as there is simply too much luck involved in scoring goals to use xGF leading to high variability between GF and xGF.









Really interesting writeup of how they developed their model here.

Feel free to talk about all things analytical here!

[butterfly]

The more I think about it, the more I wish there was a positional adjustment for wingers and centers compared to grouping them as "forwards". Wingers are the least valuable skaters and need to produce at a higher rate than centers to have value.

Some forwards can play both positions but this isn't a problem in baseball analysis as almost everyone can play multiple positions in that game.

[butterfly]

I'm not sure PK is an actual skill. Does anyone think so? If you look at the Evolving Hockey model it's almost nothing.

[Enoch Root]

Originally Posted by butterfly

I'm not sure PK is an actual skill. Does anyone think so? If you look at the Evolving Hockey model it's almost nothing.

If you do it more, would you expect to get better at it? If so, it's a skill.

[butterfly]

Originally Posted by Enoch Root

If you do it more, would you expect to get better at it? If so, it's a skill.

I’m not sure. The Sharks had a top PK for a couple years when they were terrible and it doesn’t seem to matter much in the Evolving Hockey model. They’re still terrible, I know.

[PaperBagger'14]

Being a rested team against a team on a B2B is more important than luck.

[butterfly]

Originally Posted by PaperBagger'14

Being a rested team against a team on a B2B is more important than luck.

Being a rested team would seem to improve μ according to the article SuperMatt posted in the Colorado PGT. As does being the home team. However σ is so large that I’d choose luck over any other advantage.

[PaperBagger'14]

Ok can you show me a commonly held definition for “luck” in terms of statistics? Without that settled it’s just discussing a vague platitude.

Personally I’m going with the quantifiable stat of win (or point) percentage of rested vs tired teams.

https://www.mckeenshockey.com/nhl-bl...stedtired-b2b/

[butterfly]

Originally Posted by PaperBagger'14

Ok can you show me a commonly held definition for “luck” in terms of statistics? Without that settled it’s just discussing a vague platitude.

Personally I’m going with the quantifiable stat of win (or point) percentage of rested vs tired teams.

https://www.mckeenshockey.com/nhl-bl...stedtired-b2b/

Pretty much whenever variance is in your favor. Like if you sit there gambling for an hour repeating the same bet and end up ahead of the expected outcome by any amount.

This is a good example of Baltimore having their runs distributed in almost an optimal pattern, which isn’t skill. It’s luck. They won a ton of close games and when they lost it was a blowout - exactly how you’d want your runs (or goals) distributed.

https://blogs.fangraphs.com/the-orio...dom-variation/

[Enoch Root]

I wonder which one is more repeatable though. Hmm

[PaperBagger'14]

Is that really the definition of luck? That definition misses a whole bunch of important factors. By your definition if I buy a lotto ticket for $10 and win $20 my luck would be equal to buying a $10 lotto ticket and winning $10 000 000. Each win is one win above expected.

[butterfly]

Originally Posted by Enoch Root

I wonder which one is more repeatable though. Hmm

Are you referring to gambling, PK, or the 2012 Baltimore Orioles? Sorry, I’m kind of lost.

[butterfly]

Originally Posted by PaperBagger'14

Is that really the definition of luck? That definition misses a whole bunch of important factors. By your definition if I buy a lotto ticket for $10 and win $20 my luck would be equal to buying a $10 lotto ticket and winning $10 000 000. Each win is one win above expected.

In each of those instances you are lucky. The latter is far more lucky than the former given the distribution of lottery payouts.