Hockey Analytics

[Weave]

OK, I'll play along. (I'm not actually so virulently anti-analytics as it might seem, I just tend to hate what Corsi has become.) What about an old-fashioned stopwatch? Measure possession by how long the puck is in either offensive zone, no matter who is controlling it.

 

I would be more than surprised if teams haven't tried this some time ago as a first blush look at analytics.  This is the kind of things that interns end up doing and reporting on.

I've actually been thinking that they should do this once they get more into tracking the puck movements and player movements. Track how much time the luck was in all three zones for each game.

 

The league itself may benefit most from this sort of analysis.  Neutral zone play is a yawner.  They could use this to come up with ideas to reduce play in the neutral zone in an effort to create more scoring chances.

[PASabreFan]

Why should a team be credited with possession when they don't have the puck? The whole point of measuring possession is to know who is controlling the play, and what players are best at it. I don't see a particularly good reason to "penalize" a team that is simply regrouping in its own end during a line change or whatever.

Tell that to Volchenkov.

The NHL formally embraces #fancystats

 

http://www.theglobeandmail.com/sports/hockey/revamped-nhl-website-to-track-enhanced-stats/article23107412/

 

Move along, folks.

I don't like the "us vs. them" "we won" approach. After all, the league has been keeping the stats that are turned into Corsi for decades. Adding them up and calling them something, OK whatever. The chips are the game-changer here.

[That Aud Smell]

The league itself may benefit most from this sort of analysis.  Neutral zone play is a yawner.  They could use this to come up with ideas to reduce play in the neutral zone in an effort to create more scoring chances.

 

Hmmmm. Me likey.

 

I don't like the "us vs. them" "we won" approach. After all, the league has been keeping the stats that are turned into Corsi for decades. Adding them up and calling them something, OK whatever. The chips are the game-changer here.

Okay, okay. But the "stats are stoopid" tone and vibe from others (not you, I don't think) is what brings it on.

 

And I agree: The chips are gonna be some next-level sh!t.

 

Tell that to Volchenkov.

 

Oooooh, that is such a great memory. Thanks.

Edited February 20, 2015 by That Aud Smell

[PASabreFan]

Hmmmm. Me likey.

 

Okay, okay. But the "stats are stoopid" tone and vibe from others (not you, I don't think) is what brings it on.

 

And I agree: The chips are gonna be some next-level sh!t.

 

 

Oooooh, that is such a great memory. Thanks.

 No, you're right, I think they're quite stoopid.

[TrueBlueGED]

I don't like the "us vs. them" "we won" approach. After all, the league has been keeping the stats that are turned into Corsi for decades. Adding them up and calling them something, OK whatever. The chips are the game-changer here.

If by "for decades" you mean 2007, then sure. But there's a reason nobody has put these things together for earlier years.

 

It honestly seems like half your beef is that the people who created the metrics are getting credit where you don't think it's deserved. That's fine, but you strike me as having a hard time separating that from the usefulness of the measures, and that's no bueno (or whatever Google translate says it is in Russian).

 

Last point: yes the stats are simple in nature and "discovering" them isn't exactly like Newton inventing calculus. But these guys still found new applications and new meaning to it, and that is very much credit worthy. Hell, a huge part of science is finding new ways to apply existing techniques and data.

Edited February 20, 2015 by TrueBluePhD

[That Aud Smell]

 No, you're right, I think they're quite stoopid.

that's no bueno (or whatever Google translate says it is in Russian).

 

I've noticed the Rooskie-un-ayshun of names spreading around here like an outbreak of gym mat herpes.

 

I missed the memo. What's the deal?

[Claude_Verret]

I've noticed the Rooskie-un-ayshun of names spreading around here like an outbreak of gym mat herpes.

 

I missed the memo. What's the deal?

 

Imagine the that aud smell after we trounced those dirty russki's in '76.

[Swedesessed]

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.

 

 

I do agree for the most part. It just illustrates nothing is full proof...scouts that watch players in person and get a feel for how the player is can sometimes be right, and sometimes be wrong....similar to that of analytics. In both cases, the teams that are right more then not are most successful. 

[SwampD]

If by "for decades" you mean 2007, then sure. But there's a reason nobody has put these things together for earlier years.

 

It honestly seems like half your beef is that the people who created the metrics are getting credit where you don't think it's deserved. That's fine, but you strike me as having a hard time separating that from the usefulness of the measures, and that's no bueno (or whatever Google translate says it is in Russian).

 

Last point: yes the stats are simple in nature and "discovering" them isn't exactly like Newton inventing calculus. But these guys still found new applications and new meaning to it, and that is very much credit worthy. Hell, a huge part of science is finding new ways to apply existing techniques and data.

I"m totally fine with fancystats being used to help make my team better. Unfortunately, up to now, they've only been used to tell me that my really crappy team or crappy player "isn't really that bad, here's why..."

 

I think another reason I don't really welcome them with open arms is that, well, I don't really have to have a reason. It's sports and I watch to see human beings compete, and overcome adversity,... all the usual stuff. Stats takes the humanness out of the game. It's one of the reasons why I can't f'n stand baseball, with the changing lefty/righty pitchers halfway through an inning because this guy hits blah blah bla,... ugh. 

 

With these chips, the data is only going to get even more precise, and predictions more accurate (the Madden Super Bowl was just as "exciting" as the real one with the exact same score). Maybe we'll end up just watching those in the future. At least then we don't have to worry about guys struggling with PCS

[TrueBlueGED]

I"m totally fine with fancystats being used to help make my team better. Unfortunately, up to now, they've only been used to tell me that my really crappy team or crappy player "isn't really that bad, here's why..."

 

I think another reason I don't really welcome them with open arms is that, well, I don't really have to have a reason. It's sports and I watch to see human beings compete, and overcome adversity,... all the usual stuff. Stats takes the humanness out of the game. It's one of the reasons why I can't f'n stand baseball, with the changing lefty/righty pitchers halfway through an inning because this guy hits blah blah bla,... ugh.

 

With these chips, the data is only going to get even more precise, and predictions more accurate (the Madden Super Bowl was just as "exciting" as the real one with the exact same score). Maybe we'll end up just watching those in the future. At least then we don't have to worry about guys struggling with PCS

The first paragraph simply is not true. That might be what you've encountered, but they have absolutely been used to confirm things we knew: Crosby is the best, Bergeron is the best defensive forward, Chara is amazing, and you want to have the puck and put more pressure on the opposition than they put on you. Personally I don't find it especially interesting to discuss those sorts of confirmations...I think it's much better to discuss instances where the stats and the eyes don't match up, because then there's actually a debate to be had and varying perspectives to dissect.

 

As to the rest of what you said, I can respect that from a fan perspective. We're here to be entertained after all, and if a lot of stats makes your eyes glaze over or you really don't have time to learn about them then hey, do what you want and what's enjoyable. My beef is (and maybe I'm misreading) that most anti-stats arguments aren't of this nature, but rather argue the stats don't work and have no place in a blood and guts sport fueled by passions and so on. The whole "games aren't won on a spreadsheet" and the like just drive me up a wall, especially when voiced by those who obviously don't know the stats well if at all.

[PASabreFan]

If by "for decades" you mean 2007, then sure. But there's a reason nobody has put these things together for earlier years.

1997? If so, that's three decades. :)

[Bknotz]

We always talk and hear about offensive/defensive "advanced" stats such as Corsi, but does anyone know of any advance stats that describe goaltenders? I think it would be interesting to see where some goalies on bad teams ( such as Neuvirth) stand compared to goalies on good defensive teams.

Edited February 21, 2015 by Bknotz

[rakish]

We always talk and hear about offensive/defensive "advanced" stats such as Corsi, but does anyone know of any advance stats that describe goaltenders? I think it would be interesting to see where some goalies on bad teams ( such as Neuvirth) stand compared to goalies on good defensive teams.

I wrote this a few years ago.  I was trying to get a sense of what available goaltenders at the time should be paid. I took everybody's shots against at each distance, multiplied by the league average for that distance, then combined the distances to give you an idea of how many goals they saved. Then plotted that against how much money they made.  This was before I realized that you couldn't just draw in your own line of regression because it looked good.

 

As I wrote then, I think the idea is a failure because Cody Hodgson was on the ice for worse 10 foot shots than Steve Ott was, or he was really unlucky.  This makes sense if you watched Buffalo play hockey that year, but how do you adjust for that?  The same would be true of Neuvirth in your question.  Yeah I can tell you how many 10 foot shots he sees, and what percent the average goaltender saves, but like Hodgson and Ott, is he seeing a more difficult 10 foot shot than Rinne is? I think he is, making the piece useless.

 

So I wouldn't bother reading the piece.

[X. Benedict]

Nhl.com just added "enhanced stats" ...... Not sure why they didn't just call them Corsi and Fenwick....but the best I can tell that's exactly what they are.

 

As with their inclusion in broadcasts around the league today, they look like they are here to stay.

[PASabreFan]

Nhl.com just added "enhanced stats" ...... Not sure why they didn't just call them Corsi and Fenwick....but the best I can tell that's exactly what they are.

 

As with their inclusion in broadcasts around the league today, they look like they are here to stay.

They're descriptive labels. They're trying to edumucate folks like me.

 

Yes, shot attempts is Corsi and unblocked shot attempts is Fenwick.

 

http://www.nhl.com/ice/news.htm?id=754260

[X. Benedict]

They're descriptive labels. They're trying to edumucate folks like me.

 

Yes, shot attempts is Corsi and unblocked shot attempts is Fenwick.

 

http://www.nhl.com/ice/news.htm?id=754260

Spasibo.

[SwampD]

The first paragraph simply is not true. That might be what you've encountered, but they have absolutely been used to confirm things we knew: Crosby is the best, Bergeron is the best defensive forward, Chara is amazing, and you want to have the puck and put more pressure on the opposition than they put on you. Personally I don't find it especially interesting to discuss those sorts of confirmations...I think it's much better to discuss instances where the stats and the eyes don't match up, because then there's actually a debate to be had and varying perspectives to dissect.

I just meant in relation to the Sabres. Don't you remember, we had two top twenty centers on the team. No need to make changes.:wallbash:

As to the rest of what you said, I can respect that from a fan perspective. We're here to be entertained after all, and if a lot of stats makes your eyes glaze over or you really don't have time to learn about them then hey, do what you want and what's enjoyable. My beef is (and maybe I'm misreading) that most anti-stats arguments aren't of this nature, but rather argue the stats don't work and have no place in a blood and guts sport fueled by passions and so on. The whole "games aren't won on a spreadsheet" and the like just drive me up a wall, especially when voiced by those who obviously don't know the stats well if at all.

That is real, though, and you can't dismiss it just because there is no stat for it,... yet. I guess we need something like a "Nilan quotient". If we knew we had a guy on our team with a 16.4 NQ (17.2 being the highest), and there would have been an 87% likelyhood that he would have broken Lucic's jaw in the Millier incident, our worlds could coexist. Edited February 21, 2015 by BolotoO

[carpandean]

That is real, though, and you can't dismiss it just because there is no stat for it,... yet. I guess we need something like a "Nilan quotient". If we knew we had a guy on our team who had a 16.4 NQ (17.2 being the highest), and there would have been an 87% likelyhood that he would have broken Lucic's jaw in the Millier incident, our worlds could coexist.

 

Sounds like a joint venture between nhl.com and hockeyfights.com.  Some combination of size (height, weight), hits, roughing minors, fighting majors, strength of opponents, average fight-card rating, etc.

 

Then, of course, you have to figure out what correlates to winning: a team's top rated NQ players, the average (over the roster or by TOI) NQ rating, the interaction of NQ and Corsi/Fenwick relative (a.k.a., the Gordie Howe Effect), ...

 

 

ETA: 7,777th post!

Edited February 21, 2015 by carpandean

[TrueBlueGED]

I just meant in relation to the Sabres. Don't you remember, we had two top twenty centers on the team. No need to make changes.:wallbash:

 

That is real, though, and you can't dismiss it just because there is no stat for it,... yet. I guess we need something like a "Nilan quotient". If we knew we had a guy on our team with a 16.4 NQ (17.2 being the highest), and there would have been an 87% likelyhood that he would have broken Lucic's jaw in the Millier incident, our worlds could coexist.

First, fair enough. I often wonder what Roy and Connolly would have looked like if we had Corsi and contextual stats for those first two post-lockout seasons. I have a feeling we'd have been able to see statistically what we all knew visually. Although to this day they're still the best centers by far this team has had since Drury and Briere left. You're all welcome for the reminder :p

 

As to your second point, I don't disagree at all. As always the eternal debate is the relative importance of each factor. I think a "lazy" player who puts up 50 points yearly can have a role on my team, but if you put together 12 of those guys you're probably going to have issues.

[carpandean]

They need to change the wording for SAT Rel on NHL.com:

 

Shot attempts Rel = Shot attempts for player - Shot attempts for team when player is not on the ice

 

Unfortunately (confusingly), they have a stat, SAT For, and another stat, SAT.  The former is only shots attempts by the player's team while the player is on the ice; the latter is the net shot attempts (player's team SA - opponent's SA) while the player is on the ice.  SAT Rel is based on the SAT stat, not the SAT For stat.  In other words, "Shot attempts for player" is actually the SAT stat (net shots attempted) while player is on the ice, not the SAT For stat for the player, and likewise for the team.

 

The funny thing is that the Sabres' players dominate this stat, since it subtracts off one really big negative number from a smaller negative number.  For example, Deslauriers has a STAT of -309, but leads the league with a STAT Rel of +712.  In other words, while there have been 309 more shot attempts against than shot attempts for while he was on the ice, there were actually 1,021 more shots against than shots for while he wasn't on the ice.  Seems pretty useless to not scale each by TOI or something like that.

Edited February 21, 2015 by carpandean

[TrueBlueGED]

Given that the NHL thinks zone starts and primary assists are some sort of proprietary innovation, I'm not even remotely surprised at the number of errors and oversights present.

 

Regarding the relative stats...they were always pretty weak at league-wide comparisons, to the point where I think they're basically useless outside of comparing players on the same team. Maybe adjust for team possession to make them more comparable? Don't think TOI would do the trick.

[rakish]

I have an analogy for the problem of using corsi. I'm going to ask a few questions of X, he's being generous enough to try to work through it with me, allowing me the space to get my point across. It will take a bit to get to hockey, but eventually we will get to hockey.

In the 18th century or so, the King of England decided to implement the window tax. The assessor would drive up to the house, count your windows, and levy your tax. The first two questions are, Why didn't this work? and What does it teach us about data?

[X. Benedict]

I have an analogy for the problem of using corsi. I'm going to ask a few questions of X, he's being generous enough to try to work through it with me, allowing me the space to get my point across. It will take a bit to get to hockey, but eventually we will get to hockey.

 

In the 18th century or so, the King of England decided to implement the window tax. The assessor would drive up to the house, count your windows, and levy your tax. The first two questions are, Why didn't this work? and What does it teach us about data?

 

I'll try my best.. 

 

In the 18th century when the Kind of England taxed windows, human nature being human nature, people began to brick up their windows. I'm imagining they even built houses without windows to save a few pence on the tax. Ultimately, in response to the tax, people would have hid, boarded up, or concealed their wealth by building homes with fewer, less, or probably in the case of the Irish and Scots, homes with no windows. 

 

The second part, I believe, is if the counting of windows is used by the State to determine wealth, the data, that is, the counting of windows would create a weird model. Even though the windows might  indicate some degree of wealth, they really wouldn't represent wealth. The Tax-man would have a helluva time trying to figure out where real income resided based on the measure of only windows. And daily reports on CNBC of the National Window Index would have nothing more to do with real incomes than the number of walls or faucets in any particular bungalow. 

 

How am I doing?

Edited February 22, 2015 by X. Benedict

[carpandean]

The second part, I believe, is if the counting of windows is used by the State to determine wealth, the data, that is, the counting of windows would create a weird model. Even though the windows might  indicate some degree of wealth, they really wouldn't represent wealth. The Tax-man would have a helluva time trying to figure out where real income resided based on the measure of only windows. And daily reports on CNBC of the National Window Index would have nothing more to do with real incomes than the number of walls or faucets in any particular bungalow.

 

I was thinking that the second part is basically ... you can't measure a system without affecting it and, thus, tainting your measurements.

[PASabreFan]

The Observer Effect. Also, the peasants would have been very envious when the tax man drove up in his county SUV.