Current State of Analytics in the NHL

[Doctor No]

The use of bad analytics - like Corsi - as a player evaluation tool.

How do you know that the Oilers were using Corsi as an evaluation tool?

How do you know that they didn't have the right conclusions, and didn't suffer the problems mentioned in post #15?

And supposing that they were, how do you ascribe a sample size of n=1 to failure?

 

[Mad Brills*]

Leafs have a whole analytics department as long with an assistant gm who is a huge advocate for them.

 

[Ogopogo*]

How do you know that the Oilers were using Corsi as an evaluation tool?

How do you know that they didn't have the right conclusions, and didn't suffer the problems mentioned in post #15?

And supposing that they were, how do you ascribe a sample size of n=1 to failure?

They hired a Corsi proponent and Corsi is useless at measuring individuals. Their W/L record tells the entire hockey world that what they are doing does not work.

 

[Doctor No]

They hired a Corsi proponent and Corsi is useless at measuring individuals.

Suppose that Dellow is only good at one thing (using Corsi to predict which players will do well).

It would be remarkably stupid for the Oilers to have hired him in that case, since this is something that's publicly available and easy to replicate. Preumably, Dellow offered the ability to do multiple things, and things that aren't in the public domain.

However, if your theory is true, then I agree that it would explain a lot about the Oilers and their lack of recent success.

Their W/L record tells the entire hockey world that what they are doing does not work.

Sure. Uh huh.

Are we going in circles? Because it seems like my response (that you just responded to) would work well here.

 

[Ogopogo*]

Suppose that Dellow is only good at one thing (using Corsi to predict which players will do well).

It would be remarkably stupid for the Oilers to have hired him in that case, since this is something that's publicly available and easy to replicate. Preumably, Dellow offered the ability to do multiple things, and things that aren't in the public domain.

However, if your theory is true, then I agree that it would explain a lot about the Oilers and their lack of recent success.



Sure. Uh huh.

Are we going in circles? Because it seems like my response (that you just responded to) would work well here.

If someone is a big proponent of Corsi that tells me their judgement is off. That being the case, I would have trouble trusting anything he is preaching.

 

[Doctor No]

If someone is a big proponent of Corsi that tells me their judgement is off. That being the case, I would have trouble trusting anything he is preaching.

My point is that how do you know *what* Dellow was doing with the Oilers?

And how do you know that it's a Corsi problem, and not (1) a Dellow problem, (2) a management ignoring Dellow problem, or (3) something that may have worked in the long term but wasn't showing short term results?

I've seen evidence to suggest all three.

 

[Slats432]

If the Oilers really have sent Tyler Dellow back home, we know the Oilers have failed.

QFT. If you have an analytics guy start work with a 30th place club that has decades of nepotism, bad drafting and horrible development, chances are, in four months they are still a 30th place club.

Ken Holland and Scotty Bowman at their best could have willed this team to 28th...but that is about it.

 

[CBlake]

Suppose that Dellow is only good at one thing (using Corsi to predict which players will do well).

It would be remarkably stupid for the Oilers to have hired him in that case, since this is something that's publicly available and easy to replicate. Preumably, Dellow offered the ability to do multiple things, and things that aren't in the public domain.

However, if your theory is true, then I agree that it would explain a lot about the Oilers and their lack of recent success.



Sure. Uh huh.

Are we going in circles? Because it seems like my response (that you just responded to) would work well here.

How much if the NHL's effort with analytics is not in the public domain? Kind of surprised with the 24/7 sports media world someone has not exposed it with a good feature as I sure people have seen it.

 

[Doctor No]

How much if the NHL's effort with analytics is not in the public domain? Kind of surprised with the 24/7 sports media world someone has not exposed it with a good feature as I sure people have seen it.

Team efforts are not publicized, because once proprietary work becomes public, it can be replicated and the marginal advantage of the work disappears.

Analytics are all about finding elements that are undervalued (or overvalued) in the league - if things enter the public domain, then the arbitrage opportunities disappear.

 

[CBlake]

Team efforts are not publicized, because once proprietary work becomes public, it can be replicated and the marginal advantage of the work disappears.

Analytics are all about finding elements that are undervalued (or overvalued) in the league - if things enter the public domain, then the arbitrage opportunities disappear.

I understand you want to protect your advantages, but are you surprised that at least one teams analytics efforts have not been exposed in the media through unnamed sources?

Also how much effort is put into managing the roster on a per game basis using analytics?

 

[Doctor No]

I understand you want to protect your advantages, but are you surprised that at least one teams analytics efforts have not been exposed in the media through unnamed sources?

I haven't seen one in awhile, but the nondisclosure agreements used to have significant penalties associated with them.

Also how much effort is put into managing the roster on a per game basis using analytics?

Probably not too much; if teams are chasing game-to-game variability, then that's a losing strategy. A lot of current analytic thought is about long-term trends, not small sample size events.

(With that said, remember my hypothesis that we aren't seeing the truly proprietary stuff - maybe someone's found a way to make something work, and we just haven't seen it.)

 

[CBlake]

I haven't seen one in awhile, but the nondisclosure agreements used to have significant penalties associated with them.



Probably not too much; if teams are chasing game-to-game variability, then that's a losing strategy. A lot of current analytic thought is about long-term trends, not small sample size events.

(With that said, remember my hypothesis that we aren't seeing the truly proprietary stuff - maybe someone's found a way to make something work, and we just haven't seen it.)

With the profile analytics has taken on do you see it being a viable way for a wave of kid to break into the pro sports world? How viable a career path is it?

 

[Doctor No]

With the profile analytics has taken on do you see it being a viable way for a wave of kid to break into the pro sports world? How viable a career path is it?

I'll tell you my experience - the supply of people who want to do this are far greater than the number of opportunities, so the pay likely isn't very good. With that said, someone who could demonstrate a good return on investment (somehow) could make a living, and it'd be a fun profession where you'd meet a lot of neat people.

If it's something that you had passion for, I wouldn't ever tell you not to pursue it; life's too short to be doing uninteresting work.

I'd read this thread before doing anything too seriously:
http://hfboards.mandatory.com/showthread.php?t=570990

And there are a lot of good folks to emulate. The Hockey Prospectus (and alumni) folks are good. Michael Schuckers (who posts here) is good. There are a lot of great bloggers out there. There are others posting here who do interesting (and compelling) research.

I will say this as well - there are so many more opportunities to network and meet people now (as compared to when I was looking at this in the mid-1990s). If you have good ideas, you'll get a following. It won't be easy, for sure. I'd recommend hedging your bets.

I can share more via PM if you're interested in talking further.