Hockey Analytics

[Brawndo]

Some fun facts on GMTM and his use of analytics.

 

 

First an ESPN Article regarding Teams and their use of Analytics  

 

http://espn.go.com/espn/feature/story/_/id/12331388/the-great-analytics-rankings

 

 

Second while speaking on HH earlier he brought up WOWY or With Or Without You when speaking about Dylan Strome and Connor Mc David.  

Edited February 23, 2015 by BRAWNDO

[IKnowPhysics]

First an ESPN Article regarding Teams and their use of Analytics  

 

http://espn.go.com/espn/feature/story/_/id/12331388/the-great-analytics-rankings

 

 

Neat.  We haven't yet had a compilation of where each team sits.

Edited February 23, 2015 by Я знаю физика

[TrueBlueGED]

Some fun facts on GMTM and his use of analytics.

 

 

First an ESPN Article regarding Teams and their use of Analytics  

 

http://espn.go.com/espn/feature/story/_/id/12331388/the-great-analytics-rankings

 

 

Second while speaking on HH earlier he brought up WOWY or With Or Without You when speaking about Dylan Strome and Connor Mc David.  

 

I need to listen to this interview  :wub:

[Brawndo]

I need to listen to this interview  :wub:

 

Listen at 23:30 

 

http://sabres.nhl.com/v2/ext/2014-15%20Season/HOCKEY%20HOTLINE/20150223-TimMurray-HH.mp3

[SwampD]

So, eventually, every team will be using analytics (if every team isn't already), thus minimizing their effectiveness. How long before we get to complain about it being an unfair advantage because all the really good statisticians go to the big market teams to make more money? :devil:

[Kevbeau]

I quickly read a few posts, but I haven't read the entire thread so please excuse me if I repeat anything.

 

Unless the data recording methods have vastly improved since I was last exposed to them, the biggest issue facing the use of advanced analytics in hockey is data collection. There's a significant margin of error in the raw data that you can pull from nhl.com, so I would be very wary of doing any sort of in depth trend or explanatory analysis from it. Having said that, simple analyses (such as Corsi) can be used for directional indication. Hence the caution issued when measuring single period/game performance, but generalizations can be made with increased sample sizes...often  referred to as the "law of big numbers."

 

I don't know for sure, but I would imagine the teams advancing their capabilities are using in-game or post-game data collection techniques not associated with the NHL off-ice crews. Whether it's eyeballs on a screen or some sort of visual recording software, I'm not positive. Again, there's too much error in the generic NHL locational stats (shots, hits, give/takeaways) to offer any short term findings as definitive. TOI even has a noticeable margin of error.

[TrueBlueGED]

http://m.espn.go.com/nhl/story?storyId=12370826&src=desktop&rand=ref~%7B%22ref%22%3A%22http%3A%2F%2Ft.co%2FKvcdWmtFY5%22%7D

 

Good stuff here from Stan Bowman.

[rakish]

I quickly read a few posts, but I haven't read the entire thread so please excuse me if I repeat anything.

 

Unless the data recording methods have vastly improved since I was last exposed to them, the biggest issue facing the use of advanced analytics in hockey is data collection. There's a significant margin of error in the raw data that you can pull from nhl.com, so I would be very wary of doing any sort of in depth trend or explanatory analysis from it. Having said that, simple analyses (such as Corsi) can be used for directional indication. Hence the caution issued when measuring single period/game performance, but generalizations can be made with increased sample sizes...often  referred to as the "law of big numbers."

 

I don't know for sure, but I would imagine the teams advancing their capabilities are using in-game or post-game data collection techniques not associated with the NHL off-ice crews. Whether it's eyeballs on a screen or some sort of visual recording software, I'm not positive. Again, there's too much error in the generic NHL locational stats (shots, hits, give/takeaways) to offer any short term findings as definitive. TOI even has a noticeable margin of error.

The amount of error in the RTSS reports doesn't bother me as much as it probably should. One thing that would have helped a lot is if they used 'line change' and an event.  You end up with 45 seconds between events, different players on the ice, who played those 45 seconds, don't know.  As you say, it evens out with enough data.

 

Yeah, you gotta believe each team serious about this stuff has a flock of interns taking notes gathering data.  And that why I got out of this as an interest, because my lack of data is insurmountable.  How do you know the difference whether a team knows what they are doing or not?  How do I know whether my work is any good?  I moved on to the draft, nice bright lines.

[TrueBlueGED]

The amount of error in the RTSS reports doesn't bother me as much as it probably should. One thing that would have helped a lot is if they used 'line change' and an event.  You end up with 45 seconds between events, different players on the ice, who played those 45 seconds, don't know.  As you say, it evens out with enough data.

 

Yeah, you gotta believe each team serious about this stuff has a flock of interns taking notes gathering data.  And that why I got out of this as an interest, because my lack of data is insurmountable.  How do you know the difference whether a team knows what they are doing or not?  How do I know whether my work is any good?  I moved on to the draft, nice bright lines.

 

Kind of tangental to what you're saying, but this is why I don't put much stock in the argument that the competitive advantage will evaporate once everyone is using the same data. The data might eventually be the same, but the analyses it's used in and the interpretations most certainly will not be.

[rakish]

Kind of tangental to what you're saying, but this is why I don't put much stock in the argument that the competitive advantage will evaporate once everyone is using the same data. The data might eventually be the same, but the analyses it's used in and the interpretations most certainly will not be.

Absolutely, I think the results from drafting are way more consistent than you do because some teams are using relevant analysis and most teams aren't, but I can measure these things working in the draft, but how much does data matter in the NHL? I have no idea.  I don't know what data people are looking at, or any idea how many games are won or lost because of it.

[IKnowPhysics]

http://m.espn.go.com/nhl/story?storyId=12370826&src=desktop&rand=ref~%7B%22ref%22%3A%22http%3A%2F%2Ft.co%2FKvcdWmtFY5%22%7D

 

Good stuff here from Stan Bowman.

 

Hmm. I know sports management guys like to keep proprietary secrets, but it really seems as if he believes that if he shares, he loses.

[That Aud Smell]

http://m.espn.go.com/nhl/story?storyId=12370826&src=desktop&rand=ref~%7B%22ref%22%3A%22http%3A%2F%2Ft.co%2FKvcdWmtFY5%22%7D

 

Good stuff here from Stan Bowman.

 

Stick tap to Ghost for having essentially said this (see below) many times. The distinction that Bowman draws is that everyone is not yet looking at the same things. Again: Not yet.

 

"Bowman: Once anything is widely adopted, there's no advantage. Then you're all looking at all the same things the same way. Right now, I think we're looking at things slightly differently. I don't get into specifics. It's not to be coy; I'm not trying to get accolades."

[TrueBlueGED]

Stick tap to Ghost for having essentially said this (see below) many times. The distinction that Bowman draws is that everyone is not yet looking at the same things. Again: Not yet.

 

"Bowman: Once anything is widely adopted, there's no advantage. Then you're all looking at all the same things the same way. Right now, I think we're looking at things slightly differently. I don't get into specifics. It's not to be coy; I'm not trying to get accolades."

 

I actually think the most important part of that quote is "same way." Especially once SportsVU is widely adopted and the data available, everybody is going to be looking at the same thing. But in no way, shape, or form, does that mean they'll be looking at it the same way--and in my not so humble opinion, what you do with the data is infinitely more important than the equivalence of availability.

[LGR4GM]

I thought I might put this here. It breaks down the unsustainable shooting percentages of a group of players in the league. Basically it is foreshadowing who is overproducing.  

 

 

Below, you'll find a data table with an emphasis on shooting percentages. What we know, from years and years of data, is that it is unlikely that a player can maintain a shooting percentage that deviates significantly from his career numbers for long amounts of time (which translates to lots of shots).The aim of this weekly article is to highlight players, that the data suggests, are playing at levels inconsistent with their actual talent. The last column (Dev) measures the deviation from career levels. 

http://leftwinglock.com/articles.php?id=2491&title=Unsustainable-Skaters-March-4-2015

[Taro T]

I actually think the most important part of that quote is "same way." Especially once SportsVU is widely adopted and the data available, everybody is going to be looking at the same thing. But in no way, shape, or form, does that mean they'll be looking at it the same way--and in my not so humble opinion, what you do with the data is infinitely more important than the equivalence of availability.

Not sure how the league tracks the RTSS, but the teams track their own events very precisely and at least some of them have been doing it for 10+ years now. They've manually tracked when guys hit the ice, when they come off, who has the puck and what was done with it in realtime. What they've done w/ that data definitely varies from team to team.

 

The SportsVU should increase the accuracy of the data obviously. I'd be interested to see how the league makes that data available both to teams and the general public. Doubt anyone wants their opponents looking at their precise data.

[Wraith]

I actually think the most important part of that quote is "same way." Especially once SportsVU is widely adopted and the data available, everybody is going to be looking at the same thing. But in no way, shape, or form, does that mean they'll be looking at it the same way--and in my not so humble opinion, what you do with the data is infinitely more important than the equivalence of availability.

 

This is a good time to point out that when "everyone is looking at it the same way" there is an enormous opportunity for the smart and clever people to gain an advantage. That is the real premise of "Moneyball." "Moneyball" was not about adherence to one particular school of analytical thought. It was about identifying the market inefficiencies and exploiting them. When everyone (the "market") is on the same page about what is valuable and not valuable, it becomes very cost effective to find the hidden value in the "not valuable." Sooner or later, everyone else will recognize that hidden value and the "not valuables" become "valuable." In the process, formerly "valuables" will be "not valuables" and the process will repeat. The system is read and react, and therefore will never reach an equilibrium and will be entertainingly dynamic for those not afraid of change.

Edited March 4, 2015 by Wraith

[LGR4GM]

This is a good time to point out that when "everyone is looking at it the same way" there is an enormous opportunity for the smart and clever people to gain an advantage. That is the real premise of "Moneyball." "Moneyball" was not about adherence to one particular school of analytical thought. It was about identifying the market inefficiencies and exploiting them. When everyone (the "market") is on the same page about what is valuable and not valuable, it becomes very cost effective to find the hidden value in the "not valuable." Sooner or later, everyone else will recognize that hidden value and the "not valuables" become "valuable." In the process, formerly "valuables" will be "not valuables" and the process will repeat. The system is read and react, and therefore will never reach an equilibrium and will be entertainingly dynamic for those not afraid of change.

I agree.  For instance the perfect hockey example is everyone can look at shooting percentage and yet year after year after year the GM's around the league pay guys based on outlier years when their shot % was sometimes double the norms...

[TrueBlueGED]

Not sure how the league tracks the RTSS, but the teams track their own events very precisely and at least some of them have been doing it for 10+ years now. They've manually tracked when guys hit the ice, when they come off, who has the puck and what was done with it in realtime. What they've done w/ that data definitely varies from team to team.

 

The SportsVU should increase the accuracy of the data obviously. I'd be interested to see how the league makes that data available both to teams and the general public. Doubt anyone wants their opponents looking at their precise data.

Pretty sure it was reported the NHL plans on paywalling the data a la the NFL's all-22.

[Taro T]

Pretty sure it was reported the NHL plans on paywalling the data a la the NFL's all-22.

The raw data or what they decide to aggregate out of it?

[TrueBlueGED]

The raw data or what they decide to aggregate out of it?

That's an excellent question to which I have no answer. I could have missed it, but from what I remember there weren't any specifics reported, only that the NHL plans to monetize the data.

[Taro T]

That's an excellent question to which I have no answer. I could have missed it, but from what I remember there weren't any specifics reported, only that the NHL plans to monetize the data.

My guess is that they'll have 1 level available to teams about their own players, another available leaguewide to the teams and those 'in the loop', and a lowered level available to the public. Would be pretty cool to have access to the raw data.

[BagBoy]

Looks like Ottawa is taking analytics with a grain of salt.

 

http://www.tsn.ca/patrick-wiercioch-analytics-vs-eyeball-test-1.223118

[dudacek]

Looks like Ottawa is taking analytics with a grain of salt.

 

http://www.tsn.ca/patrick-wiercioch-analytics-vs-eyeball-test-1.223118

 

What an interesting article. 

I wonder, could you substitute Tim Murray for Bryan Murray and Mark Pysyk for Patrick Weircoch?

 

Or maybe Weircoch was a Tim Murray favourite. We still don't have a real could handle on what kind of players Tim likes, because so many of his acquisitions haven't had a chance to show their stuff yet.

Edited March 6, 2015 by dudacek

[TrueBlueGED]

If you don't believe Corsi has any value, you should probably ignore what I'm about to post. Otherwise, I think it's interesting. 

 

Corey Pronman ‏@coreypronman  23m23 minutes ago

In the NHL at even-strength, forwards define a team's puck posession. Role determined by ESTOI/60 team rank

 

Corey Pronman ‏@coreypronman  21m21 minutes ago

Analytics help you predict performance, but also helps us understand the game. Value of Fs over Dmen have been my major eye-opener this yr.

 

Corey Pronman ‏@coreypronman  6m6 minutes ago

I discussed this at the start of the season, but am of the firm belief that a top 6 is much more important than a top 4. See previous tweet.

 

I'm not sure how I feel about posting this, because it may contribute to a few bridge jumpers if we have to draft Hanifin instead of Eichel or McDavid :lol:

 

In seriousness though, I do think it's interesting how, at least this year, the forwards have been really driving the bus. I'd really like to see this analysis over a several year spread to see how consistent these results are.

[qwksndmonster]

Is that each player's individual corsi?  Or team corsi%?