Hockey Analytics

[BagBoy]

Let’s talk “analytics”, shall we? As will become obvious to proponents of it, I’m with those who think it’s complete horseshlt. I am biased against it, but I’m not inflexible. After all I was initially shocked in a bad way about this Kane trade today. But now, after reading all of the opinions here today, and reflecting objectively, I feel great about it! So for all you proponents of analytics, convince us doubters!

 

First of all, this analytics business in hockey sure seemed to spring up suspiciously right after the Moneyball movie. I thought it was a good movie (and factual), and I think analytics from a baseball perspective makes perfect sense, and I believe the data supports that. I’m an Orioles fan, and I think Earl Weaver was a genius. He was the first guy to track specific pitcher vs. hitter stats back in the early 70’s (on hand-written index cards, not computers btw) and he used them brilliantly. But here’s the thing…hockey isn’t baseball.

 

My understanding of the Corsi is that at its core, it is simply trying to quantify puck possession, since there is nobody out there tabulating the actual time of possession, like they do in soccer. Maybe it’s time to start tabulating time of possession. I think that would be a very informative stat. Anyway the Corsi is obviously a flawed way of quantifying possession. For example, those Soviet teams of the 70’s and 80’s virtually always had an obvious advantage in time of possession over Team Canada. They did this by doing precisely the opposite of what the Corsi rewards, namely being very selective with their shots, and passing the puck back and forth until they got a good shot opportunity. So, if we’re trying to quantify puck possession, don’t you think counting passes would be a much more valid piece of the puzzle than how many times you got your shot blocked? I sure do.

 

Basically all it is is the SOG plus missed shots and blocked shots, so this means the more times you get your shots blocked or you miss the net, the “better” you are. So therefore, if a player comes down the wing and shoots wide causing the puck to go all the way back to his own zone, he still gets a “plus”. Or how about when a D-man gets the puck at the point, has time to wrist it toward the net, but chooses instead to crank a full wind up, and then the opposing winger blocks the shot and creates a breakaway for himself? The D-man gets a Corsi “plus” for this.

 

I don’t doubt for a second that higher Corsi numbers have and always will correlate positively with better teams. I’m not saying the stat is “wrong” per se. It’s just a stupid stat. Teams with the highest ‘goals for’ vs. ‘goals against’ ratios also tend to be the best teams, but nobody gets excited about that stat.

 

Plain old-fashioned shots on goal (SOG) is a far superior stat to the Corsi, and the only reason the Corsi appears to have any relevance is because SOG’s are the vast majority of its data points. Every SOG “could” be a goal. Zero missed nets and almost zero blocked shots can be goals. Missing the net is very frequently a bad play for the offense, and blocked shots are almost never good for an offense. These are not stats that should be rewarded. No coach of mine ever gave me a pat on the back for missing the net or getting a shot blocked.

 

I realize I haven’t addressed the iCorsi stuff or the Fenwick stuff, or any of the rest of it, because admittedly I have dismissed them all as more of the same.

If any of that is actually worthwhile, I need to be illuminated.

Please discuss.

[Claude_Verret]

Posters like TrueBlue can offer more insight, but I found this article to give a good introduction to many of the basic hockey fancystats concepts.

 

http://www.thehockeynews.com/blog/thn-analytics-an-introduction/

 

I think PA brought up the issue of shot counting bias in different rinks, and it sounds like there is some work going on there among the stat geeks.
 

 

These five over-arching concepts are some of the more-important ideas to emerge from the modern hockey stats movement, but hockey stats continue to evolve. To date, adjustments to Corsi measures are being tested (a scoring chance study revealed that scoring chances and Corsi run extremely close together), the value of tracking the way teams enter the zone is becoming more widely accepted, and a variety of adjustments are used to address rink counting bias. As analytics continue to push towards conclusions like the ones above, and expand on the principles of Sinden, Neilson, and Barnes, we only get closer to a better understanding of the game.

 

 

 

[IKnowPhysics]

I could do some heavy lifting on this, but at the moment, I'm lazy.  So here's the lazy answer:  GMs use them and they say they work.

 

Advanced statistics have been around for years.  The story of how they got their real traction in baseball is romanticized in Moneyball, but the same revolution occurred for some teams in hockey roughly 2-5 years after the events of Moneyball took place.  Not every team/GM used them in some form or another, but almost all teams do now.  Even old timers like Ken Hitchcock examine advanced stats after every period.  Some NHL teams actively employ full-time stats analysts.  The Sabres' Jason Nightingale, Neil McKenney, and Graham Beamish all do focused analytics work.  It's likely that the others scouts, AGMs, and GMTM all have a competent understanding of the fundamentals of hockey analytics.  GMTM and Kevin Devine may very well be experts at this.

 

 

 

“Statistics are like a lamp post for a drunk,” [Corsi] said. “You can either use it to illuminate or to lean on. So buyer beware. … But without a doubt it lends me some more information. If you think with your heart, you end up with heart failure.”

 

 

 

“The players wanted the line together,” [Hitchcock] said. “We put it together thinking it was going to work. The data after 10 games said, get away from this line. It doesn’t work. Get away from it quickly because it’s a disaster. Three games later I finally stopped putting the line together because the data was off the charts bad. But visually, I thought it works. It didn’t get backed up by data and the analytical information was right on the mark. And it was stuff, I guarantee if I mention the names, you’d say that’s a great line. It should work. We’ve seen it before and (the line) was a nightmare.”

 

How do you analyze hockey?  How do you reveal useful aspects of the game that are buried?  That may contradict a coach's intuition?  The NHL is pretty bad at gathering useful statistics on a nightly basis.  I think that'll change moving forward, but for now, there's only a few reliable stats you can use to unearth information about quality of play.  The most useful ones available currently are based on shots.

 

 

 

Plain old-fashioned shots on goal (SOG) is a far superior stat to the Corsi, and the only reason the Corsi appears to have any relevance is because SOG’s are the vast majority of its data points. Every SOG “could” be a goal. Zero missed nets and almost zero blocked shots can be goals. Missing the net is very frequently a bad play for the offense, and blocked shots are almost never good for an offense. These are not stats that should be rewarded. No coach of mine ever gave me a pat on the back for missing the net or getting a shot blocked.

Now, when you analyze the usefulness of a stat, statements like this are of zero value.  None.  That's caveman thinking.  The usefulness of a particular statistic or advanced statistic is determined by more analysis and math: how well it correlates to success or losses; scoring chances for or against; goals or goals against; wins or losses.  Calculations determine whether a stat is useful.  Statistical methods, although complicated, are used to develop new tools and quantities to analyze.  It doesn't matter if it sounds like , as long as it has been mathematically shown with sound methods to not be .

 

Advanced statistics and analytics are an attempt at scientifically approaching hockey questions that were previously thought of as quantitative:  How good is player x?  Was it because of his teammates or opponents, or was he really that good?  How do you compare players on different teams?  These aren't easy questions to tackle with math, but information can be and is revealed this way.

 

Now, as Corsi's quote suggests, advanced stats should be taken with a grain of salt, because they do have limits.  They have very specific meanings and limitations and they should always be understood in context.

[TrueBlueGED]

I'm pretty tired of arguing about the viability of these things, and I'm certainly not going to put much of any effort into arguing with somebody who has summarily dismissed them in advance. The war is over, the nerds won. Neither you nor anybody else has to be happy about it, but it doesn't change reality. Some of the most successful NHL teams are using statistics to elevate their franchises (Chicago and LA being the two most successful with it) and they have access to proprietary stats that we don't...but that doesn't change that what we do have available has enhanced our understanding of the game and ability to evaluate players for those who care how to use the stuff properly. And that's really the key, is using it properly--knowing what the stats are really saying, knowing their limitation, and thinking about the implications the right way.

 

Use it, don't use it, like it, don't like it...whatever. But the nerds did win, and at some point people will start accepting it.

Edited February 13, 2015 by TrueBluePhD

[SwampD]

It's not Hockey analytics, but I heard today that the show House of Cards was created through the use of analytics. The number crunchers even mentioned the name Keven Spacey. Kinda creepy.

[LastPommerFan]

It's not Hockey analytics, but I heard today that the show House of Cards was created through the use of analytics. The number crunchers even mentioned the name Keven Spacey. Kinda creepy.

Absolutely. Unlike traditional delivery methods, netflix has the ability to parse the minute by minute viewing habits of it's customers. Think of how powerful that can be. I hope Terry's Analytics department finally comes to the conclusion that what we really want are Stanley's and Lombardi's. I imagine they'll screw something up and will end up with an old Kubrick screenplay starring Vince Vaughn. 

[I am Defecting]

If is fixed data, such as pitcher vs batter, "advanced" dumb-country-bumpkin stats are very useful (for the local economy [5 jobs at least].)  5 vs 5 stats contain too many variables, and any hardcore believer in Professor Corsi or Mrs.Fenwick is "highly irrational."  Tee-heee.  I do Star-Trek.  

 

I like time of possession idea of OP.

[French Collection]

NHL put chips in pucks and player jerseys at All Star game. Time of Possession (TOP) stats could be tracked accurately with this type of technology. No need for stopwatches. Scrims would have to be blanked out, scrambled. I'm not a techie but I can see possibilities with that stat.

Back to the Russian teams of the past, they were generally outshot by Canadian teams but probably had the puck 60% of the time. Canada scored by shot volume and Russians scored by shot quality.

[SwampD]

Absolutely. Unlike traditional delivery methods, netflix has the ability to parse the minute by minute viewing habits of it's customers. Think of how powerful that can be. I hope Terry's Analytics department finally comes to the conclusion that what we really want are Stanley's and Lombardi's. I imagine they'll screw something up and will end up with an old Kubrick screenplay starring Vince Vaughn. 

In terms of the arts, I think it is the most horrifying idea ever. My wife and I have a term for it when it comes to music, "lowest common denominator plus one".

 

I also wonder if analytics would have told Hasek that he needed to reduce all of those low percentage tendencies that he somehow made work, thereby neutering him.

[LastPommerFan]

In terms of the arts, I think it is the most horrifying idea ever. My wife and I have a term for it when it comes to music, "lowest common denominator plus one".

 

I also wonder if analytics would have told Hasek that he needed to reduce all of those low percentage tendencies that he somehow made work, thereby neutering him.

House of Cards is far from LCD+1, so maybe netflix is doing something different with their numbers.

 

Analytics would have made Hasek the highest paid athlete in the world. His statistical advantage was enormous during his era.

[I am Defecting]

Hasek was good, but I saw it with my own eyes, and that's irrelevant in this era of "advanced" statistics that have dominated my hockey team, and radio (coincidentally, cause they're not only the OFFICIAL station, they're also IMPARTIAL)

 

Now, I think that you should trade away your best goalie, because great goalies are a dime a dozen.  Just look at Neuvirth.  He only lost a measly single point that Enroth would have won, so far.

 

That's insignificant.  Nowadays, we're too smart for Hasek types.  We're smart fans.  We use advanced statistics.  Good goalies are caveman .

[...]

Wasn't there a start up thread like this a couple of weeks ago? Seems that way after a 6 day absinthe.

 

 

I also wonder if analytics would have told Hasek that he needed to reduce all of those low percentage tendencies that he somehow made work, thereby neutering him.

 

Not that I have anything against the stats, I do think they have an obvious place as a significant tool for the craftsman, but this is an interesting question.  With the heavy usage of stats now, would they get in the way of a natural talent whose ability goes against the statistical norm?

[SwampD]

Speaking of stats and goalies, did anyone else see the HNIC breakdown of Millers game during one of the intermissions of the Canucks/Flames game on Saturday (Hockey Day, or Canada Day, or whatever it was called)? It was really cool and I can't find it anywhere.

 

It pretty much said we've all said, and there may be somethin to Ghost's theory. From inside where most shots come from, Miller is 1st among all goalies. From all the perimeter zones (I think there were six) he was 29th and worse.

 

They then went on to show the subtle changes in style over time. Good stuff. I wish I could find it.

[Stoner]

Speaking of stats and goalies, did anyone else see the HNIC breakdown of Millers game during one of the intermissions of the Canucks/Flames game on Saturday (Hockey Day, or Canada Day, or whatever it was called)? It was really cool and I can't find it anywhere.

 

It pretty much said we've all said, and there may be somethin to Ghost's theory. From inside where most shots come from, Miller is 1st among all goalies. From all the perimeter zones (I think there were six) he was 29th and worse.

 

They then went on to show the subtle changes in style over time. Good stuff. I wish I could find it.

One of the zones was "He never saw it," the other was "Never had a chance." A third zone was curiously labeled "WTF!"

[SwampD]

House of Cards is far from LCD+1, so maybe netflix is doing something different with their numbers.

 

Analytics would have made Hasek the highest paid athlete in the world. His statistical advantage was enormous during his era.

I'm guessing that Netflix'ssss stats guys would tell you that that is exactly what it is, and they couldn't be happier.

[TrueBlueGED]

In terms of the arts, I think it is the most horrifying idea ever. My wife and I have a term for it when it comes to music, "lowest common denominator plus one".

 

I also wonder if analytics would have told Hasek that he needed to reduce all of those low percentage tendencies that he somehow made work, thereby neutering him.

I'm not aware of stats for goaltender technique, so I'm gonna go with no.

[That Aud Smell]

GMs use them and they say they work.

The war is over

 

This thread's, like, a year or so late.

 

Seriously.

 

I started out quite skeptical about #fancystats when it comes to the fluid, immeasurably intangible, scrap-a-licious game of hawkey. 

 

But, like TrueBlue said: The war is over. There's no debate left to have. Those who continue to engage in such a debate are just diddling their ding-a-ling, or maybe diddling their neighbour's.

 

The thing that threw for a funking loop was when a client (executive type) sent me a service contract to review for a vendor who -- wait for it -- is in the business of providing advanced statistics for business managers. These are metrics used to indicate not just productivity and the like, but also worker happiness, morale, etc.

 

The future is here. Like it or don't.

[SwampD]

This thread's, like, a year or so late.

 

Seriously.

 

I started out quite skeptical about #fancystats when it comes to the fluid, immeasurably intangible, scrap-a-licious game of hawkey. 

 

But, like TrueBlue said: The war is over. There's no debate left to have. Those who continue to engage in such a debate are just diddling their ding-a-ling, or maybe diddling their neighbour's.

 

The thing that threw for a funking loop was when a client (executive type) sent me a service contract to review for a vendor who -- wait for it -- is in the business of providing advanced statistics for business managers. These are metrics used to indicate not just productivity and the like, but also worker happiness, morale, etc.

 

The future is here. Like it or don't.

Maybe that is why sports has become so boring.

 

I guess all I have to look forward to is the day when analytics finally comes up with a stat that shows that the other team always knows exactly what you are going to do, so you have to be unpredictable and go against what analylics tells you to do 47% of the time in order achieve success.

[That Aud Smell]

Maybe that is why sports has become so boring.

 

I guess all I have to look forward to is the day when analytics finally comes up with a stat that shows that the other team always knows exactly what you are going to do, so you have to be unpredictable and go against what analylics tells you to do 47% of the time in order achieve success.

Have sports become boring? I'm asking.

 

The NFL had a pretty horrendous year. How're things in the NHL? I must admit: With the Sabres being awful, I'm not aware of how the league is more generally.

 

I think that the numbers are just a different way of understanding, and verifying, what you're seeing. I don't think that competitive sports will become boring because of #fancystats, and I don't think #fancystats will ever drive game-play in the way you are suggesting.

[Stoner]

There's no debate left to have. Those who continue to engage in such a debate are just diddling their ding-a-ling, or maybe diddling their neighbour's.

 

Regarding possession, there is room for debate. Reasonable people can debate how to measure possession. Is it best measured by shots directed at the goal? That's not why Corsi came up with it. And it took some dweeb in Edmonton hearing Darcy talking about Corsi's stat to come up with the idea for measuring possession that way. Seems fairly Scotch-taped together. Then, the kicker is that the data is suspect. Everyone knows that. What's the value of the product of a shaky philosophy and bad data?

 

I'm looking forward to the NHL chip thingie so we can have a precise idea of possession time, so the nerds can go back to heating up Hot Pockets and watching War Games. (Running out of snark.)

[SwampD]

Have sports become boring? I'm asking.

 

The NFL had a pretty horrendous year. How're things in the NHL? I must admit: With the Sabres being awful, I'm not aware of how the league is more generally.

 

I think that the numbers are just a different way of understanding, and verifying, what you're seeing. I don't think that competitive sports will become boring because of #fancystats, and I don't think #fancystats will ever drive game-play in the way you are suggesting.

I'm pretty sure it already is. Hasn't it been responsible for the players that the Sabres decided to keep and the ones they let go for the past 16 years?

 

The day when the players we ice or, more importantly, the art we produce is being determined by an algorithm, that's the day you might as well just send the HKs to my house to put a bullet in my head, because forget about Judgement Day, Skynet is already in control.

[TrueBlueGED]

Regarding possession, there is room for debate. Reasonable people can debate how to measure possession. Is it best measured by shots directed at the goal? That's not why Corsi came up with it. And it took some dweeb in Edmonton hearing Darcy talking about Corsi's stat to come up with the idea for measuring possession that way. Seems fairly Scotch-taped together. Then, the kicker is that the data is suspect. Everyone knows that. What's the value of the product of a shaky philosophy and bad data?

 

I'm looking forward to the NHL chip thingie so we can have a precise idea of possession time, so the nerds can go back to heating up Hot Pockets and watching War Games. (Running out of snark.)

There is no debate. Corsi and Fenwick are proxies for possession, not a direct measure...but they're still valid and incredibly useful. In an effort to appeal to your sensibilities, it passes the eye test. When the Sabres, a terrible Corsi team, plays a team like Detroit, a great Corsi team, the Sabres sure look like they never have the puck. When evenly matched Corsi teams like Chicago and Nashville play, the game appears balanced and highly competitive. The numbers bear out what the eyes see, or vice versa depending on your preference. There is no debate.

[That Aud Smell]

Regarding possession, there is room for debate. Reasonable people can debate how to measure possession. Is it best measured by shots directed at the goal? That's not why Corsi came up with it. And it took some dweeb in Edmonton hearing Darcy talking about Corsi's stat to come up with the idea for measuring possession that way. Seems fairly Scotch-taped together. Then, the kicker is that the data is suspect. Everyone knows that. What's the value of the product of a shaky philosophy and bad data?

 

I'm looking forward to the NHL chip thingie so we can have a precise idea of possession time, so the nerds can go back to heating up Hot Pockets and watching War Games. (Running out of snark.)

All good points. The actual creation of the data is still quite rudimentary, even flawed. It still has real value, though. When the chip-tracking comes, the data will have still more value.

I'm pretty sure it already is. Hasn't it been responsible for the players that the Sabres decided to keep and the ones they let go for the past 16 years?

 

The day when the players we ice or, more importantly, the art we produce is being determined by an algorithm, that's the day you might as well just send the HKs to my house to put a bullet in my head, because forget about Judgement Day, Skynet is already in control.

What you're talking about is roster-construction, not game-play.

 

Now, #fancystats may well inform a coach's decision on which lines or players to send over the boards in certain situations, but #fancystats are not going to dictate what happens as between and among the 12 (or fewer) guys on the ice at any given time -- rather, those fellas are going to dictate and create what the #fancystats are.

 

So, feel free to ease up on the apocalypse talk. Sports are still fun.

[TrueBlueGED]

I'm pretty sure it already is. Hasn't it been responsible for the players that the Sabres decided to keep and the ones they let go for the past 16 years?

 

The day when the players we ice or, more importantly, the art we produce is being determined by an algorithm, that's the day you might as well just send the HKs to my house to put a bullet in my head, because forget about Judgement Day, Skynet is already in control.

The numbers don't tell anyone to do anything by themselves, but you know this. I suppose I'll never understand the rejection of additional information when making decisions. It's no different than house shopping and wanting to know of you're looking in a flood zone. Do you just assume because the day of your walk through happens to be sunny that it will always be that way? Or do you want the historical data too?

[Claude_Verret]

Speaking of stats and goalies, did anyone else see the HNIC breakdown of Millers game during one of the intermissions of the Canucks/Flames game on Saturday (Hockey Day, or Canada Day, or whatever it was called)? It was really cool and I can't find it anywhere.

 

It pretty much said we've all said, and there may be somethin to Ghost's theory. From inside where most shots come from, Miller is 1st among all goalies. From all the perimeter zones (I think there were six) he was 29th and worse.

 

They then went on to show the subtle changes in style over time. Good stuff. I wish I could find it.

 

 

There very well may be something with Miller's game that causes his struggles with shots from farther out, but to me Drane's impaired depth perception theory doesn't hold water simply because that problem should have the reverse effect in terms of shot distance.  As someone who is legally blind in one eye  I can tell you that if you lightly toss me a baseball from five feet away, it will be a crapshoot as to whether or not I'll catch it.  But throw me a heater from 60'6" and I'll catch it every single time.