Is there an equivalent of a "Moneyball" for the NHL?

[SJGoalie32]

I agree with some others that I don't believe hockey is a sport that can be easily subjected to statistical modeling.

Baseball is pretty easy because, as Moneyball and Sabremetrics point out, it's a game of individual battles that occur 50-100 times per game, in every game, 162 times per year. The individual strengths of each batter/pitcher combination, fielders, stadium dimensions, etc. can vary, but over the course of a season (or multiple seasons), they tend to balance each other out. There are minor differences, but every at-bat is pretty much the same. Every shift in hockey is different.

Look at a guy like Jonathan Cheechoo. His goal totals went: 9, 28, 56, 37, 23, 12. You just don't see those kind of massive swings in home runs or batting average in baseball unless a guy misses half the season. Cheech has been relatively banged up, but he's played in no less than 80% of the games in each season. Granted most NHL players don't have swings that massive either, but they seem far more common than in baseball.

Each at-bat in baseball seems to have a (relatively) very narrow set of outcomes. A very narrow set of variables (0-3 runners on base, 0-2 outs, etc.). Hockey has so many variables, so many outcomes, it would really seem just too many to calculate. And the variables which have such a seemingly minor influence in baseball (a good pitcher will deflate a good hitters stats, but he will deflate ALL the hitters stats about the same on average), have a far greater impact in a hockey game.

*************

One other thing to consider, too....

As Billy Beane said (and was pointed out in an earlier post), the playoffs are basically a crap shoot. You can predict the outcome of players over the course of an entire season where all the variables tend to even out, but in a short playoff series, the variables have too much influence to the point where success becomes impossible to predict.

That's not too a big deal for a baseball team. You play 162 games, and then the best 8 teams out of 30 (roughly speaking) get into the playoffs. The goal of a GM is to be one of those top 8 teams. From there, you need to win 11 games over 3 series. But in the NHL, the playoffs expand to 16 teams and 4 sets of best-of-7. Being the best team in the division/conference/league becomes almost irrelevant. In most seasons, just having a .500 record is enough to get your foot in the door and a chance at the Cup, and from there the game of chance takes over.

In the NHL, being the absolute best regular season team only nets you a Cup like 30% of the time. That's a pretty low percentage. Even if there was some magical formula you could implement that GUARANTEED that you would ice the best team in the league each and every year.....that would only be good for a championship 3 out of every 10 seasons. Statistically speaking.

 

[Enstrom39]

I agree with some others that I don't believe hockey is a sport that can be easily subjected to statistical modeling.

Baseball is pretty easy because, as Moneyball and Sabremetrics point out, it's a game of individual battles that occur 50-100 times per game, in every game, 162 times per year. The individual strengths of each batter/pitcher combination, fielders, stadium dimensions, etc. can vary, but over the course of a season (or multiple seasons), they tend to balance each other out. There are minor differences, but every at-bat is pretty much the same. Every shift in hockey is different.

Look at a guy like Jonathan Cheechoo. His goal totals went: 9, 28, 56, 37, 23, 12. You just don't see those kind of massive swings in home runs or batting average in baseball unless a guy misses half the season. Cheech has been relatively banged up, but he's played in no less than 80% of the games in each season. Granted most NHL players don't have swings that massive either, but they seem far more common than in baseball.

I'm pretty sure I could find some batters with some pretty wild swings in their HR numbers. How about Brady Anderson's HR totals?

I would also point out that in baseball each player's stats are skewed by park effects. In hockey there are no park effects to deal with in the modern since all rinks are the same dimensions.

Again, the question isn't "can you model hockey perfectly?" the question is "can modeling hockey help you improve your decision making?" and I think the answer is "yes" to the 2nd question.

 

[Patman]

I'm pretty sure I could find some batters with some pretty wild swings in their HR numbers. How about Brady Anderson's HR totals?

I would also point out that in baseball each player's stats are skewed by park effects. In hockey there are no park effects to deal with in the modern since all rinks are the same dimensions.

Again, the question isn't "can you model hockey perfectly?" the question is "can modeling hockey help you improve your decision making?" and I think the answer is "yes" to the 2nd question.

Exactly.

The idea isn't "well, can i find a 50 goal scorer" but rather "can I guess who is going to be a 50 goal scorer better than the other guy?"

The A's, the subject of Moneyball, wasn't trying to isolate the next 5 .320 30HR guys but they were trying to isolate those that were more likely to get there. Granted, no tool is really going to get you that but often it can be just as valuable as eyeballing a player. Similarly, they were able to isolate the ideal qualities that they could glean out of players (On-Base as a superior measure to Batting Avg or Slugging). They had an idea of how to read the statistical part of baseball better and they had an idea of how to identify talent.

Unfortunately, I don't think we can address the latter due to lack of data on those leagues not influenced by the NHL (I think some good work could be done with the AHL if they have the same equipment do what the NHL currently does). College baseball stats are everywhere.

I do think that the game itself can be re-understood given the right tools. I think current NHL talent that is undervalued can be found.

There are teams that must be wallowing in cash. My dream job would be, personally, to have a go at being a mathematical statistician for say the Leafs or the Rags. I say those teams because I have to imagine they have more money to spare on a lark such as a small stat analysis group. Especially with the Leafs history of floundering as they have. I'm shocked that with their potential resources that they haven't branched out. Surely that cup is worth an extra 500K USD/yr. for 4-5 statistical researchers.

 

[Enstrom39]

There are teams that must be wallowing in cash. My dream job would be, personally, to have a go at being a mathematical statistician for say the Leafs or the Rags. I say those teams because I have to imagine they have more money to spare on a lark such as a small stat analysis group. Especially with the Leafs history of floundering as they have. I'm shocked that with their potential resources that they haven't branched out. Surely that cup is worth an extra 500K USD/yr. for 4-5 statistical researchers.

Ah, but the big problem is that NHL GMs are almost all former players or scouting types who inherently trust their own eyes more than relying on numbers.

Plus there is a huge risk for any organization that breaks from the pack and takes a new route. Much easier to CYA going the traditional route.

That said from it seems that it is the west coast teams (SJS, LAK, VAN) that are the most inclined to use stats. Gillis is a former agent and therefore probably more comfortable with numbers. Sharks are in the heart of hi tech R&D and that environment might inform their management style. I don't know much about the Kings situation.

 

[pitseleh]

That said from it seems that it is the west coast teams (SJS, LAK, VAN) that are the most inclined to use stats. Gillis is a former agent and therefore probably more comfortable with numbers. Sharks are in the heart of hi tech R&D and that environment might inform their management style. I don't know much about the Kings situation.

Lombardi is a former player agent too and was San Jose's GM until 2003, so he may have carried over what he learned/setup in San Jose to the Kings.

 

[TheRumble]

I agree that hockey isn't as easily statistically modeled as baseball but I don't think that the fundamental argument of Moneyball is in stats. Rather, the book is about finding undervalued assets.

If you look at the method used by Billy Beane in Moneyball, you'll see that most teams have adapted similar methods with great success and now it's all over the league so the strategy is no longer as effective. If you look at the Oakland A's roster this year though, you'll find that they have a lot of players with proven track records (Giambi, Garciappara, Chavez) but have been discounted by other teams because of their injury histories so the philosophy is still the same, just the method is different. That said, the A's are in last place in their division so maybe that method isn't very effective.

I think hockey is more comparable to football in that a player's effectiveness is somewhat dependent on the type of system they play in. Teams like the Colts and Patriots always seem to find impact players despite picking late in the draft while perennial losers like the Lions have a decade worth of busts despite picking in the top 10 almost every year.

 

[Jumptheshark]

Because hockey is more of team sport where you have more rolling action. as oposed to baseball where when I have gone to some games--A left fielder did not get a ball hit to him for the full 9 innings---hockey is always rolling and all players on the feild is always in motion

 

[marc-edouard grier]

I think hockey is more comparable to football in that a player's effectiveness is somewhat dependent on the type of system they play in. Teams like the Colts and Patriots always seem to find impact players despite picking late in the draft while perennial losers like the Lions have a decade worth of busts despite picking in the top 10 almost every year.

Two articles on why that might be the case:

http://www.footballoutsiders.com/walkthrough/2009/walkthrough-made-not-born

http://brodeurisafraud.blogspot.com/2009/03/development-is-key-not-drafting.html

 

[CanadaBacon]

that seems like a useless stat for anyone but penalty killers. The player in the box has no effect on the penalty kill, the goalie's performance, or the PP's performance unless that player is a penalty killer.

You dont see how taking a penalty and putting your team down a man in the first place can affect a game? really?

 

[pitseleh]

Two articles on why that might be the case:

http://www.footballoutsiders.com/walkthrough/2009/walkthrough-made-not-born

http://brodeurisafraud.blogspot.com/2009/03/development-is-key-not-drafting.html

Freakonomics had an interesting article about it lately:

http://freakonomics.blogs.nytimes.com/2009/04/22/is-the-top-nfl-draft-pick-a-penalty/

 

[marc-edouard grier]

Freakonomics had an interesting article about it lately:

http://freakonomics.blogs.nytimes.com/2009/04/22/is-the-top-nfl-draft-pick-a-penalty/

Thanks for the heads up . Rings extremely true to me, considering I'm a Raiders fan (sigh), and most of those thoughts went through my head in the 07 draft.

Reminds me of Moneyball, where Beane intentionally picked some of "his" guys in the first round, partly because he wouldn't have to pay those guys top dollar, considering the weaker bargaining power the prospects he chose held.

 

[golfortennis]

I agree that hockey isn't as easily statistically modeled as baseball but I don't think that the fundamental argument of Moneyball is in stats. Rather, the book is about finding undervalued assets.

If you look at the method used by Billy Beane in Moneyball, you'll see that most teams have adapted similar methods with great success and now it's all over the league so the strategy is no longer as effective. If you look at the Oakland A's roster this year though, you'll find that they have a lot of players with proven track records (Giambi, Garciappara, Chavez) but have been discounted by other teams because of their injury histories so the philosophy is still the same, just the method is different. That said, the A's are in last place in their division so maybe that method isn't very effective.

I think hockey is more comparable to football in that a player's effectiveness is somewhat dependent on the type of system they play in. Teams like the Colts and Patriots always seem to find impact players despite picking late in the draft while perennial losers like the Lions have a decade worth of busts despite picking in the top 10 almost every year.

I would say the Red Sox are the best example of Moneyball with money to spend. They look for the same things the A's do, except that they can pay for them if they have higher prices.

The problem isn't that the A's philosophy doesn't work, it's that the assets they seek have become valuable on the marketplace and have moved beyond their price range. Plus, it is nearly impossible for Billy Beane to make a trade anymore because if he calls asking for a player, the other team is going to up its asking price automatically.

I don't think it's so much system in the case of the football teams. The Colts start with the best QB in the game, and work from there. Manning makes so many players better because he can thread a needle. The Patriots are similar in that they start with Tom Brady, who was a real good college QB. He didn't "meet" the metrics for a high draft pick, he can just play football. Both teams have struggled when good players have been hurt.

I would liken the Denver Broncos under Mike Shanahan more towards what you were saying. Because of their cut blocking schemes, they can put basically anyone back there and have a 1,000 yard rusher.

 

[Mathletic]

Two articles on why that might be the case:

http://www.footballoutsiders.com/walkthrough/2009/walkthrough-made-not-born

http://brodeurisafraud.blogspot.com/2009/03/development-is-key-not-drafting.html

I partly disagree with the article. Although I believe in teachers as head coach and assistants to develop players and implementing a sound coherent system, I've found patterns that lead me to believe that good players aren't all made from the day they're drafted. A lot of the players who were so called late bloomers like Duncan Keith, Martin St Louis and whoever have patterns that are closely related to those of higher draft picks who are thought of as sure shots. Also, I've found that a lot of busts could have been avoided from Marcel Hossa to Jason Ward ... I'm a habs fan ... to whoever. I can't predict all busts yet like Jason Bonsignore and Alexandre Daigle because I haven't looked in all directions yet due to lack of time, but I've cleared a whole bunch of them only based on their years before being drafted.

 

[Mathletic]

Thanks for the heads up . Rings extremely true to me, considering I'm a Raiders fan (sigh), and most of those thoughts went through my head in the 07 draft.

Reminds me of Moneyball, where Beane intentionally picked some of "his" guys in the first round, partly because he wouldn't have to pay those guys top dollar, considering the weaker bargaining power the prospects he chose held.

poor you, it's a thing a beauty watching the Patriots work the draft and trade down when they need and making great trades adding future picks year after year, in the meantime, you see the Raiders reaching for players round after round and year after year. It's fine that they draft Heyward-Bay, but please, work the draft and make it go your way, you can get him mid to late first, get extra picks in the process.

I don't believe much in the strategy of getting your guy as soon as you can. I'm much more in favor of the Patriots strategy to move down, get your guy about where he's about to go and get a bunch of picks in the process. Also, I don't believe the Brady pick was as random as people make it seem. Again, I don't believe the Patriots reach for players at any point in the draft, and although they might not have thought he'd become a super bowl mvp, I'm sure the pick wasn't random, just like the Cassell and O'Connell pick weren't random late picks.

 

[Mathletic]

anyone knows what's up with Alan Ryder from hockeyanalytics? I sent him an email and he hasn't answered back

also, what's Mike Smith's company that does some stats work for the NHL? former Hawks GM

 

[LadyStanley]

http://www.canada.com/sports/flex+their+intelligence+draft+later+rounds/1738809/story.html

One proponent of "Moneyball" like usage for the NHL has analyzed the draft tendencies over the last 30 years. How you draft in 3rd and later rounds really shows how good a team is.

(So the fact that Nabokov is a 9th round pick, Pavelski a 7th round is one reason why the Sharks are one of two teams he ranks with A+.)

 

[Wetcoaster]

When Mike Gillis took over the GM position of the Canucks, he said from the beginning that he was going to closely follow the "moneyball" strategy in his management. He never went into any real detail, but he's taken the team into a playoff position so we'll see how it pays off.

Here is an article about analytics and how it can be used in the NHL. As noted in the article former GM Mike Smith and a partner have started a company to provide these services (Coleman Analytics) and reportedly have 5 NHL clients as of May 2008. Smith points out it is hard to get the NHL to accept new ways of doing things.
http://www2.canada.com/vancouversun...=3b895917-da15-4828-a14c-13f093dbbc0b&k=74598

Also see:
http://www.sportsbusinessjournal.com/article/53651

 

[mucker*]

No....and fans of the NHL should be happy there is no moneyball.

This is one of the most over-hyped, garbage information out there.

It's true some of the money ball priniciapls are useful....things such as On-base-percentage.

However moneyball does not believe in stealing bases and the recommendations it makes for not having a closer shows the authors/creators of this theory have no regard at all for the physical nature of pitching and no common sense for sports.

It took a few novel ideads, like on base percentage and scouting, and added garbage like no stealing and no set 'closer', and was nothing more than a clearinghouse for stat geeks who never played the sport.


How you can recommend that it is wise to take out a dimension of the game, like stealing bases, really is ridicolous.
Ask any reliever to how realistic it is not to have 'set' roles.

The theory is worthless and the Red Sox have won because of good scouting, money, and the OBP.
They dismissed the closer theory and they do steal bases.


The A's always choked in the playoffs because they never could manufacture a run.

 

[Dado]

The A's always choked in the playoffs because they never could manufacture a run.

This is pretty unfair. The As did extremely well based on their limited financial resources, and continue to do so. As recently as the '08 season, they have an exceptional win record for the dollars that they are able to spend.

That teams like the Red Sox do better isn't because they don't do a form of moneyball, it's that they do moneyball with an extremely high level of resources. Smarts + money generally trumps smarts + no-money.

However moneyball does not believe in stealing bases and the recommendations it makes for not having a closer...

These are both incorrect statements.

Sabremetrics *likes* stealing bases - if you can do it well enough. That means not being overly aggressive - only send the guys most likely to make it and only against pitchers/catchers less likely to get the out and only do it in ballparks where the runs-scored expectation tilts it in your favor.

Nor does it dislike closers - what it dislikes is spending so much money on a closer that it takes away your ability to hire quality starters. Same deal with relievers. You can see this with the Red Sox - the money their closer makes is FAR less than the top-paid closers earn. If Papelbon ends up wanting the big paycheck, it is a very safe bet Boston will trade him rather than pony up the $12-15M he'd likely fetch on the open market, because that difference will buy you a very good starter.

 

[Enstrom39]

Here is an article about analytics and how it can be used in the NHL. As noted in the article former GM Mike Smith and a partner have started a company to provide these services (Coleman Analytics) and reportedly have 5 NHL clients as of May 2008. Smith points out it is hard to get the NHL to accept new ways of doing things.
http://www2.canada.com/vancouversun...=3b895917-da15-4828-a14c-13f093dbbc0b&k=74598

Also see:
http://www.sportsbusinessjournal.com/article/53651

I'm very skeptical of the public descriptions of the Coleman Analytics. So far it sounds like a bunch of "clutch scoring" which is probably just small sample size theater.

I looked at Game Winning Goals years ago and they were largely random across a players career, the only thing that jumped out is that players know for their assist had a higher than expected proportion of GW goals--my hypothesis is that while they usually favor the pass, with the game on the line they are more likely to break with their pattern and take a shot, and therefore more likely to end up with a GW Goal. It is also possible that opposing goalie and defenders expect them to pass and thus are more vulnerable if an Adam Oates choose to shoot instead of pass late in a game.

There is a lot of interesting stuff being done on repeatable skills like Corsi numbers and so forth.

 

[Mathletic]

No....and fans of the NHL should be happy there is no moneyball.

This is one of the most over-hyped, garbage information out there.

It's true some of the money ball priniciapls are useful....things such as On-base-percentage.

However moneyball does not believe in stealing bases and the recommendations it makes for not having a closer shows the authors/creators of this theory have no regard at all for the physical nature of pitching and no common sense for sports.

It took a few novel ideads, like on base percentage and scouting, and added garbage like no stealing and no set 'closer', and was nothing more than a clearinghouse for stat geeks who never played the sport.


How you can recommend that it is wise to take out a dimension of the game, like stealing bases, really is ridicolous.
Ask any reliever to how realistic it is not to have 'set' roles.

The theory is worthless and the Red Sox have won because of good scouting, money, and the OBP.
They dismissed the closer theory and they do steal bases.


The A's always choked in the playoffs because they never could manufacture a run.

rest assured, the NHL is still in stone ages when it comes to statistical analysis, I tried to forward my model to certain NHL teams and came across certain professors who tried to advance their work also, and it took them a lot of calls and time before having very little interest. It seems like certain teams are getting interested but it's nothing to write home about.

apparently video work is still only in its infancy in the NHL. I thought players spent most of their day in the video room when they were done training and practicing, but apparently not. You see the NFL players have rooms in the stadium so they can sleep there at night when they're done with the video and can start the next day early morning, but apparently in hockey it's still far removed from that. Imagine where analytics are if video is still basically new to them.

 

[Blue Dragon]

I think one of the things which could be really important in such an analysis is Penalties Taken. Obviously fighting majors would be excluded from such an analysis. Any guy who can get the other teams down a man for 2 minutes even if they have less skill in other areas of the game is doing something very valuable for their team imo.

 

[ColumbusNucksFan]

Have you guys done any work in game theory? I was trying to use game theory along with statistical analysis to look at shootouts in hockey for one of my graduate stats class.

 

[Mathletic]

Have you guys done any work in game theory? I was trying to use game theory along with statistical analysis to look at shootouts in hockey for one of my graduate stats class.

I have but mainly from an administrative pov. I read a basketball book a while ago though that applied concepts of game theory to the game of basketball. I can't recall the name though, I'll try to find it back.

 

[Mayor Bee]

No....and fans of the NHL should be happy there is no moneyball.

This is one of the most over-hyped, garbage information out there.

It's true some of the money ball priniciapls are useful....things such as On-base-percentage.

However moneyball does not believe in stealing bases and the recommendations it makes for not having a closer shows the authors/creators of this theory have no regard at all for the physical nature of pitching and no common sense for sports.

I thnk you've missed the entire point of the book and the A's application of it. The central tenet is how a team that has no business winning more than 65 games a year regularly competes for playoff spots by attempting to determine exactly what type of tradeoff can be used in order to create a certain type of lineup.

Stealing bases requires a certain minimum percentage to become useful to the offnese rather than a liability. Oakland in their lengthy post-Rickey Henderson era has never had anyone able to steal at this minimum threshold, whether they were brought up through the farm system, traded for, or signed as a free agent.

As for the physical nature of pitching, there's approximately one person in the planet who knows what he's talking about. I'm not him (although I've studied his methods carefully) and neither are you.

It took a few novel ideads, like on base percentage and scouting, and added garbage like no stealing and no set 'closer', and was nothing more than a clearinghouse for stat geeks who never played the sport.

The idea of a relief ace who is used in important situations (for example, a tied game in the 7th inning) rather than in "save situations" (up three runs in the 9th with no one on base) is brilliant. It's something that was an established standard for a lot longer than the idea of a one-inning closer that has created all-stars out of Billy Koch.

As for "never played the sport", who cares. How many top-level executives in sports actually played that sport at a high level? That'd be a very, very short list.

How you can recommend that it is wise to take out a dimension of the game, like stealing bases, really is ridicolous.

See above.

Ask any reliever to how realistic it is not to have 'set' roles.

If I need to know how to construct a team, the last people I'm asking are players. We can ask Elroy Face about how detrimental not having a "set role" was, or maybe Bruce Sutter or Dan Quisenberry or Goose Gossage.

The theory is worthless and the Red Sox have won because of good scouting, money, and the OBP.
They dismissed the closer theory and they do steal bases.

Why hasn't it worked for the Yankees since 2000?

The A's always choked in the playoffs because they never could manufacture a run.

Or because a five-game series is a complete crapshoot in which an inferior team wins a decent amount of the time.