Giving some context to analytics, their role and they're not as new as you think

[joshjull]

Here is a recent series of tweets by Hawerchuk@BehindtheNet about the recent surge in analytics talk and hirings. They give a nice framework for how analytics are best used, how the hype over them recently is getting carried away and their not quite as new as you think.

Hawerchuk @behindthenet · 1d 1 day ago
As we approach the trough of disillusionment for hockey analytics, here are a few helpful thoughts…

Helpful thought #1) Hockey insiders have been using "analytics" for decades. +/-, Sinden/Corsi shot/pass/touch counting, video aggregation

these stats had the imprimatur of cigar-chomping insiders, so nobody dug too deeply or cared too much

One day, members of the general public found out what insiders had been doing and slowly worked through the value of this data

somehow people popularizing the league's internal metrics became outsiders as far as fans/media were concerned

it's a classic obtuse battle of ideas. e.g. Obamacare was conservative for Romney but socialist for Obama

it took "analytics" predicting an unavoidable Leafs collapse to push people to the "Peak of Inflated Expectations"

It's amazing - people promoted ideas the NHL used for decades w/o press caring but these ideas then needed to be proven publicly

now that analytics have been re-proven externally, teams have been getting PR boosts by announcing various hires

but teams were already using analytics. So there's no new benefit.

Except I suppose people will expect the Leafs to benefit, hence the trough of disillusionment

Now here's the missing piece: teams need to know how to interpret these stats and correctly use them to drive decisions

it's statistical parallel to @Lowetide_'s "saw him good" principle. Teams don't understand regression will cut guy based on 3 games

Teams need to take long view to get analytics benefit and incentives don't align. Today's best GMs still only have 14-day outlook

May see benefit for poorly-run teams [TOR, EDM] but the inflated expectations are that "stats guys" will take them on a 2015 playoff run

unless @mc79hockey is making $1M a year, we need to seriously temper expectations

There's very little low-hanging fruit in analytics and most of it has been harvested in hockey

There's no Matt Stairs or Roberto Petagine waiting to be freed

Click to read more...

 

[Chad Thundercock]

I agree that the 'analytics' hire has become a PR boosting thing, see the Bills for this especially. But, spare me on the usual elitist 'team insiders know everything, common public are fools' crap from Hawerchuk.

 

[joshjull]

I agree that the 'analytics' hire has become a PR boosting thing, see the Bills for this especially. But, spare me on the usual elitist 'team insiders know everything, common public are fools' crap from Hawerchuk.

Umm thats not what he said. He basically said most GMs (Insiders) don't use analytics properly because they are thinking too short term.


What he said about the public is that analytics are something new to them but not to the teams themselves. Because of this expectations are unrealistically elevated by these recent hires. That fans feel the stats guys will work wonders for their teams and thats not likely to happen for the reasons stated. Expectations need to be tempered as he put it.

 

[Husko]

Trying to read that made my head hurt

 

[Djp]

Umm thats not what he said. He basically said most GMs (Insiders) don't use analytics properly because they are thinking too short term.


What he said about the public is that analytics are something new to them but not to the teams themselves. Because of this expectations are unrealistically elevated by these recent hires. That fans feel the stats guys will work wonders for their teams and thats not likely to happen for the reasons stated. Expectations need to be tempered as he put it.

My profession is in Statistics/applied Mathematics/Operations Research--if I really wanted to I could be one of these hires.

Sometimes Analytics can go too far.

Its easy to design metrics but the big question is how reliable is what its measuring,

You may think you are creating one metric but in reality this metric you are measuring really consists of 3 pieces that are merged together than if you can find a way to measure this you can get a better measure of success/failure.

+/- is a useful stat but its also a biased measure that will favor those who play on good teams with high offensive numbers at even strength.

Would a better measure be to give + credits to those directly responsible for the goal. More reasonsibilty is a higher score than less responsibility. So a defenseman who makes a mistake with a dumb pass that is intercepted in front of the net and the player scores is the only one thatgets a -1 while all the other players get a 0 on that. A player forechecking who causes the defense to cough up the puck who didnt play the puck and thus not get an assit on the goal would get a + while those not directly involved in the goal would get a 0.

If there is an extended offensive zone pressure that leads to a goal..the onest that get a +1 would be the goal scorer, primary assit, and someone who made a big play, while the other players may get only a 1/2 or 1/4 point. On defense if there was a blown assignment then that Dman gets a -1 while those not directly involved may get only a -1/2 or -1/4

 

[Push Dr Tracksuit]

Sept. 17th Marek v Wyshynski had Steve Dangle on if you like the analytics talk

didnt we used to have an all purpose analytics thread? I'd just like to know where I can find all the stats that go into these metrics

 

[Woodhouse]

didnt we used to have an all purpose analytics thread? I'd just like to know where I can find all the stats that go into these metrics

Yeah, but IIRC it was a season-centric thread and only had a handful of pages anyway. Below is what you're looking for and that's lifted from the By The Numbers forum:

Is there a list of resources somewhere?

This is far from a complete list and will be modified as time goes on. If you have any suggested additions, please feel free to PM on of the By The Numbers forum moderators.

Analytics Websites:
Hockey Prospectus
Hockey Analytics
Behind The Net
HockeyAnalysis.com

Analytics Blogs:
Hockey Historysis
Brodeur Is A Fraud
Hockey Abstract
Hockeythink
Objective NHL
NHL Numbers

Data Sources:
Extraskater.com - Contains player and team level statistics for Corsi/Fenwick/Shots as well as scoring and quality of competition. Also includes game level detail.
Stats.hockeyanalysis.com - Player and team based stats for shots/goals as well as WOWY and HART stats. Also includes some nice charts/graphs.
Behind The Net - Contains season level statistics for qualcomp, on/off plus minute, corsi, zone stats and shooting percentages
Hockey-Reference
Hockey Summary Project
HockeyDB
Hockey Databank Project
The Goaltender Home Page
NHL Standings under various scoring systems

Tools and Software:
The R Project For Statistical Computing

 

[Jim Bob]

Hawerchuk @behindthenet · 1d 1 day ago
As we approach the trough of disillusionment for hockey analytics, here are a few helpful thoughts…

Helpful thought #1) Hockey insiders have been using "analytics" for decades. +/-, Sinden/Corsi shot/pass/touch counting, video aggregation

This is like saying that Moneyball didn't really change the way MLB teams were built because insiders knew about batting average and ERA 20 years ago.

Analytics are about getting beyond simple addition, subtraction, multiplication, and division stats and into more complex algorithms to try and drive more educated decisions.

And at the end of the day, there are a ton of ways to skin a cat. You can win a Cup with a GM and a coach who love analytics and lean on them a ton and you can win a Cup with a GM and coach who go with gutalytics.

But, it is silly to try and say that analytics and their application in the NHL is nothing new.

 

[Jim Bob]

http://sports.yahoo.com/blogs/nhl-p...sby--pierre-mcguire-s-coaching-133448672.html

3. Zach Parise's hockey strategy views

In case you thought the analytics thing was a little overblown this summer, here's Zach Parise on the efficacy of carry-in zone entries:

“I read a study this summer that showed shots generated off carrying the puck in as opposed to dumping it in, and it’s like 4-to-1. It’s not even close. I just found it so interesting because everyone’s like, ‘Forecheck, forecheck, forecheck. I get it, but you dump the puck, you have to get it back. All you’re doing is giving the puck away. I mean, it’s so hard to get it, why would you give it away?â€

Hey, attaway Zach. Remember the days when troll reporters would ask players, “What do you think of corsi or whatever?†and most would say, “They're stupid. Who cares?†Well now, Zach Parise is reading studies on zone entries in the offseason and saying his team should stop dumping the puck in most situations. What a guy.

 

[joshjull]

This is like saying that Moneyball didn't really change the way MLB teams were built because insiders knew about batting average and ERA 20 years ago.

Analytics are about getting beyond simple addition, subtraction, multiplication, and division stats and into more complex algorithms to try and drive more educated decisions.

And at the end of the day, there are a ton of ways to skin a cat. You can win a Cup with a GM and a coach who love analytics and lean on them a ton and you can win a Cup with a GM and coach who go with gutalytics.

But, it is silly to try and say that analytics and their application in the NHL is nothing new.

Actually it isn't. He's saying the stats have been around but haven't been used properly because many GMs operate in such a short time frame.

And no team has ever won the Stanley with a GM and coach who love analytics. So you can't say the underlined since it hasn't happened yet.

 

[enthusiast]

http://sports.yahoo.com/blogs/nhl-p...sby--pierre-mcguire-s-coaching-133448672.html

A nice contest to Doughty dismissing Corsi

 

[joshjull]

http://sports.yahoo.com/blogs/nhl-p...sby--pierre-mcguire-s-coaching-133448672.html

Zach may want to re-read those studies. The ratio is 2 to 1 not 4 to 1. He may also want read deeper into the studies and have a better contextual understanding of things.

This may come as a shock but teams actually try to defend zone entries. When that happens teams have basically two options (1. dump it in or 2. regroup at center ice.). Regrouping actually leads to possible neutral zone turnovers which in turn could lead to zone entries into your own zone. Its actually safer to dump the puck in than re-grouping, especially with your lesser talented lines.

Zach also doesn't seem to realize that the numbers don't fully explain that it takes a certain level of talent to be able to make these zone entries. The amount of players/lines capable of consistently pushing possession by carrying the puck is very much outnumbered by the amount that can't. And you certainly don't want your lesser skilled lines attempting to regroup at center ice. So it would make sense that they dump the puck in.

That stat also doesn't factor in that some defense first strategies focus on denying the opponent shots on goal as a priority. As opposed to trying to maximize their own scoring chances. Teams that trap rely a lot on dumping the puck. Thats because turning the puck over in the neutral zone is not an option for an effective trapping team. So they are willing to forgo offensive chances from carrying the puck in in order to ensure they can set up their trap. Offensively they rely on the counterattack off the mistakes of their opponent trying to gain their zone. They prefer to get the puck deep to make the opponent have to start as far back in their own end. This enables the trapping team to set up their system. Ironically Parise went to the Stanley Cup with a team that played this way. He now plays on a team thats much more talent overall than his Devils did. Thus they are also much better and more focused on trying to gain the zone by carrying the puck. But trapping is a great way to maximize a less talented team and it would be foolish to change that approach without the required talent.


I love when these random advanced stats get thrown out there. Dump and chase will always be a part of the game for a variety of reasons.

 

[Jim Bob]

Actually it isn't. He's saying the stats have been around but haven't been used properly because many GMs operate in such a short time frame.

Helpful thought #1) Hockey insiders have been using "analytics" for decades. +/-, Sinden/Corsi shot/pass/touch counting, video aggregation

#1 [cont] these stats had the imprimatur of cigar-chomping insiders, so nobody dug too deeply or cared too much

That's not how I read that part.

And no team has ever won the Stanley with a GM and coach who love analytics. So you can't say the underlined since it hasn't happened yet.

That's like saying nobody could say that man can walk on the moon before Neil Armstrong actually did it.

I didn't say a team had.

There is no logical argument for how a team having a GM and coach that lean heavily on analytics could never win a Cup because it's never happened before.

 

[kcoxsabresfan]

Trying to read that made my head hurt

If only there was another method of communication that didn't limit a user's output in characters... We could call it an "article" and perhaps break up the ideas contained in this article into something called "paragraphs." How absurd.

 

[joshjull]

That's not how I read that part.

If you took that alone then I get your take. But he followed later with this…..

Now here's the missing piece: teams need to know how to interpret these stats and correctly use them to drive decisions

it's statistical parallel to @Lowetide_'s "saw him good" principle. Teams don't understand regression will cut guy based on 3 games

Teams need to take long view to get analytics benefit and incentives don't align. Today's best GMs still only have 14-day outlook

That's like saying nobody could say that man can walk on the moon before Neil Armstrong actually did it.

I didn't say a team had.


There is no logical argument for how a team having a GM and coach that lean heavily on analytics could never win a Cup because it's never happened before.

I'm simply stating that until a team actually wins one in that fashion its not yet one of the ways to skin a cat (win a Cup). You feeling it could happen doesn't make it so.


There is a very logic argument to made that a coach won't win a Stanley Cup leaning heavily on analytics. They can be used by a coach to inform their ideas systemically and tactically when they set things up at the start of the year. They can be used to make adjustments later in the year after a long enough sample of games transpire. But they will not be used in game because of sample size and lack of individually based numbers to predict outcomes.


A sport like baseball is tailor made for analytics because of the enormous amount of individually based statistics. A baseball manager can make in game decisions based on these stats because they are predictive of the potential outcome for a specific player in a specific situation. Plus baseball is a slow paced game which allows managers the time to use these stats as specific situations arise. This type of situation doesn't exist in hockey. Where a coach can look at an individually based stat to predict an outcome for a specific player. I don't see that changing because of the nature of hockey. There are too many interdependent factors that influence advanced stats in hockey. Things like Corsi and Fenwick rely on shots and thats not an individual stat. Nor are they stats you want to make decisions on based on 1 or 2 periods of hockey. You need a good sized sample for those numbers to be effective and informative. I just don't see how coaches will be using advanced stats to make in game decisions.

 

[joshjull]

Trying to read that made my head hurt

If only there was another method of communication that didn't limit a user's output in characters... We could call it an "article" and perhaps break up the ideas contained in this article into something called "paragraphs." How absurd.

I tried to clean it up a little more.

 

[Jim Bob]

There is a very logic argument to made that a coach won't win a Stanley Cup leaning heavily on analytics. They can be used by a coach to inform their ideas systemically and tactically when they set things up at the start of the year. They can be used to make adjustments later in the year after a long enough sample of games transpire. But they will not be used in game because of sample size and lack of individually based numbers to predict outcomes.

A coach can lean heavily on analytics and not use them in-game.

 

[Push Dr Tracksuit]

Actually it isn't. He's saying the stats have been around but haven't been used properly because many GMs operate in such a short time frame.

And no team has ever won the Stanley with a GM and coach who love analytics. So you can't say the underlined since it hasn't happened yet.

While I understand your sentiment, It's not accurate to say that since no one has done it you can't claim someone won't do it. It's also not accurate to claim that it's true because no one has proved it false. So you 2 are gonna run around all day on this one because it's a broken argument from the start.

I. Argumentum ad Ignorantiam: (appeal to ignorance) the fallacy that a proposition is true simply on the basis that it has not been proved false or that it is false simply because it has not been proved true.