Daryl Shilling "hockey Sabremetrics"?

[Ruslan Zainullin]

So under a google search for hockey sabermetrics i came across the name Daryl Shilling. His site appears to no longer to be active and I couldn't find anything else about it. I guess my question to others on the board is, why has sabermetric type analysis not taken off in hockey? wouldn't it make the debates we all have ranking prospects better to have concrete data to say "player x is better than player y"? Or, team a's roster is better than team b's roster based on the data? wouldnt it be helpful to know how much better a third liner on the Devils is than a replacement 4th line call up from the AHL?

I know hockey isnt baseball, its far less of a situational game and far more of a reactionary sport than baseball is, but I think there would be value in this kind of in depth statistical analysis, I even think that if the data was out there it could make scouting better and help teams make better personnel decisions. what say you hf boards?

Click to read more...

 

[Shawa666]

So under a google search for hockey Sabremetrics i came across the name Daryl Shilling. His site appears to no longer to be active and I couldn't find anything else about it. I guess my question to others on the board is, why has Sabremetric type analysis not taken off in hockey? wouldn't it make the debates we all have ranking prospects better to have concrete data to say "player x is better than player y"? Or, team a's roster is better than team b's roster based on the data? wouldnt it be helpful to know how much better a third liner on the Devils is than a replacement 4th line call up from the AHL?

I know hockey isnt baseball, its far less of a situational game and far more of a reactionary sport than baseball is, but I think there would be value in this kind of in depth statistical analysis, I even think that if the data was out there it could make scouting better and help teams make better personnel decisions. what say you hf boards?

Because, baseball being a zero sum game, applying statistics to it is easier done than it is to hockey.

 

[Fourier]

So under a google search for hockey Sabremetrics i came across the name Daryl Shilling. His site appears to no longer to be active and I couldn't find anything else about it. I guess my question to others on the board is, why has Sabremetric type analysis not taken off in hockey? wouldn't it make the debates we all have ranking prospects better to have concrete data to say "player x is better than player y"? Or, team a's roster is better than team b's roster based on the data? wouldnt it be helpful to know how much better a third liner on the Devils is than a replacement 4th line call up from the AHL?

I know hockey isnt baseball, its far less of a situational game and far more of a reactionary sport than baseball is, but I think there would be value in this kind of in depth statistical analysis, I even think that if the data was out there it could make scouting better and help teams make better personnel decisions. what say you hf boards?

You may want to read some of the blogs written about the Oilers. "Advanced statistics" is a near religion amongst many. For the most part I personally think that the effort is noble but the result, in particular how the data is interpreted, is often very questionable.

Baseball is a far more discrete sport than hockey. BUt if you are looking for some info take a look here:

http://hockeyanalytics.com/

There is some interesting.

 

[TaketheCannoli]

Because, baseball being a zero sum game, applying statistics to it is easier done than it is to hockey.

Can you explain what you mean by this?

 

[Shawa666]

Can you explain what you mean by this?

Each offensive action can be tallied in some way against a defensive player, and vice versa.

 

[Sensfanman]

Hockey is very complex and hard to analyze; this is compounded by some of the worst stat tracking in any sport.

For example, if Lidstrom is backing up a 2 on 1 and denies the pass, saving a possible goal, should he be rewarded? Maybe he should take the shooter.
To know, you'd need:
1) Probability of a goal given the player with the puck scores
2) Probability of a goal given the player with the puck passes
3) Probability of Lidstrom intercepting the pass
4) Probability of the pass being missed and puck ends up along the boards
5) Probability the pass is completed
5a) Probability the player scores given the pass is completed

As well as the impact each of those events has on the outcome of the game. And this is probably missing a lot of things, such as the state of everyone else on the ice. If it's just the 3 players, that's one thing but if Datsyuk is hussling back to help, Lidstrom might change how he plays and the evaluation changes with it.

Regardless, as you can see, to measure a simple play, you need a lot of data, which leads to my next point. Hockey doesn't track possession data. We can somewhat measure #1 and 5a above, but anything else is a complete guess. To make matters worse, some of the more interesting data we have (Hits, Blocks, Giveaways, Takeaways) are measured so poorly by the NHL that they have no meaning.

The reason why stats are not used in hockey is simply because we don't have stats to use. It's really pathetic.

 

[LadyStanley]

Corsi seems to be the new metric for tracking players

http://www.sportsology.info/Conflic...tatistical-Analysis-Usefulness-and-Prevalence

Here's a blog/editorial-like piece arguing about the usefulness.

 

[Fourier]

Corsi seems to be the new metric for tracking players

http://www.sportsology.info/Conflic...tatistical-Analysis-Usefulness-and-Prevalence

Here's a blog/editorial-like piece arguing about the usefulness.

Corsi is passé LS. It is now CorsiRel.

 

[Bear of Bad News]

We do a lot of the advanced hockey analytics over in the History section - I'll come back with some links when I get a chance, but poking around over there wouldn't hurt.

 

[danishh]

In baseball 1+1= 2. In hockey 1+ 1 can equal anywhere between zero and four.

 

[KevFu]

Perfect example:

Keith Jones was once (something ridiculous like) -7 in a game. And in the post-game review of the film, the coaches pointed out that he was on his man and where he was supposed to be on every single goal.

 

[SmellOfVictory]

Corsi is passé LS. It is now CorsiRel.

Just a variant of Corsi, but yes. Realistically, CorsiRel taking into account quality of competition and zonestarts is the best measure by most opinions.

And it hasn't taken off in hockey for any number of reasons (I know a lot of mainstream writers have accused it of being too math-y and not being able to capture enough of the game), but it's definitely gained a very strong core of support amongst hockey nerds.

 

[Fourier]

Just a variant of Corsi, but yes. Realistically, CorsiRel taking into account quality of competition and zonestarts is the best measure by most opinions.

And it hasn't taken off in hockey for any number of reasons (I know a lot of mainstream writers have accused it of being too math-y and not being able to capture enough of the game), but it's definitely gained a very strong core of support amongst hockey nerds.

I've been accused on many occasions of not understanding the "math" but even though I feel I am a "hockey nerd" I don't see that much value in either Corsi or Corsi Rel, at least not the way they are typically used in arguments about individual players.

 

[Fourier]

We do a lot of the advanced hockey analytics over in the History section - I'll come back with some links when I get a chance, but poking around over there wouldn't hurt.

A lot of the stuff I have seen in the HoH board seems to revolve around finding ways to compare eras. I'd be curious to know how much stock you put into most of this. To me it seems to have issues in dealing with extremes, though it is often the extremes that people want to talk about.

 

[Bear of Bad News]

A lot of the stuff I have seen in the HoH board seems to revolve around finding ways to compare eras. I'd be curious to know how much stock you put into most of this. To me it seems to have issues in dealing with extremes, though it is often the extremes that people want to talk about.

I agree with your last sentence entirely (well, essentially entirely).

The primary reason for adjusting statistics for eras is to give a different tint on the lens with which we view players of bygone years. It's an attempt to say "how impressive was this [game/season/career]" by reframing it in metrics that we're comfortable with today.

The problem with adjusted numbers is in how they're often used. They provide additional information, but certainly not a single sum total that can be used in an unbiased fashion across eras. Adjusted numbers can't tell you (very well) how Gordie Howe would have played in an NHL with Russians, for instance. Or how Patrick Roy would have done without a facemask. Or how John Ross Roach would have fared against the slapshot. Troubles develop when people try to do too much with the new information.

Here's a recent (today) discussion in the HOH forum that is pretty relevant to this conversation:
http://hfboards.com/showthread.php?t=969063

I won't introduce all of the participants, but Iain Fyffe is one of the contributing authors of Hockey Prospectus, who do this kind of thing for a living.

 

[Fugu]

We do a lot of the advanced hockey analytics over in the History section - I'll come back with some links when I get a chance, but poking around over there wouldn't hurt.

Shoo now, back to your HOH section. No recruiting of our BOHB posters.


))

I've been accused on many occasions of not understanding the "math" but even though I feel I am a "hockey nerd" I don't see that much value in either Corsi or Corsi Rel, at least not the way they are typically used in arguments about individual players.

I agree. A course on Poisson's work might give these so-called nerds a new view on stats and hockey.

Edit: I do have a secret rule that if someone accuses Fourier of not getting the math involved, it's an instant lifetime ban.

A lot of the stuff I have seen in the HoH board seems to revolve around finding ways to compare eras. I'd be curious to know how much stock you put into most of this. To me it seems to have issues in dealing with extremes, though it is often the extremes that people want to talk about.

I honestly cannot wrap my head around the methodology. If we accept that hockey is in large part governed moreso by random events (at the game not individual level), then account for the fact that the individual variables are too great (x injured, who played with whom, who officiated, someone had a cold, the general pace of the game being different due to conditioning/training, the coaches use different methods, different rules, center lines/no center lines, and so on.....).... you cannot realistically compare anything across decades. The one argument I can accept to a certain extent, although this too may suffer from these same nefarious variables, is the relative standing amongst one's peers.

 

[Jeffrey93]

So under a google search for hockey Sabremetrics i came across the name Daryl Shilling. His site appears to no longer to be active and I couldn't find anything else about it. I guess my question to others on the board is, why has Sabremetric type analysis not taken off in hockey? wouldn't it make the debates we all have ranking prospects better to have concrete data to say "player x is better than player y"? Or, team a's roster is better than team b's roster based on the data? wouldnt it be helpful to know how much better a third liner on the Devils is than a replacement 4th line call up from the AHL?

I know hockey isnt baseball, its far less of a situational game and far more of a reactionary sport than baseball is, but I think there would be value in this kind of in depth statistical analysis, I even think that if the data was out there it could make scouting better and help teams make better personnel decisions. what say you hf boards?

Just buy EA's latest version of NHL. It's got all the data in there!

 

[Fourier]

I agree with your last sentence entirely (well, essentially entirely).

The primary reason for adjusting statistics for eras is to give a different tint on the lens with which we view players of bygone years. It's an attempt to say "how impressive was this [game/season/career]" by reframing it in metrics that we're comfortable with today.

The problem with adjusted numbers is in how they're often used. They provide additional information, but certainly not a single sum total that can be used in an unbiased fashion across eras. Adjusted numbers can't tell you (very well) how Gordie Howe would have played in an NHL with Russians, for instance. Or how Patrick Roy would have done without a facemask. Or how John Ross Roach would have fared against the slapshot. Troubles develop when people try to do too much with the new information.

Here's a recent (today) discussion in the HOH forum that is pretty relevant to this conversation:
http://hfboards.com/showthread.php?t=969063

I won't introduce all of the participants, but Iain Fyffe is one of the contributing authors of Hockey Prospectus, who do this kind of thing for a living.

The bolded part is really my big beef with most of this stuff. Another example is Desjardin's NHLE. I think that on whole it does a pretty good job of illustrating the fall-off as players move from league to league. But when one starts to argue that player X is a better prospect than player Y becuase his NHLE is 38 points vs 32 I start to wonder if people have any idea what so ever about what they are saying.

Another example was the intense debate that centered around RNH's chances of being a bust because of the relative balance between his even strength scoring and his pp points. What was ridiculous to me was that it completely ignored the player's circumstance. It also ignores, as does NHLE, that the counter examples to the collective wisdom seem to come primarily, and in large numbers, from the exact group to which the method is so often applied, high-end prospects. (This is in essence the similar of trying to use smoothing techniques to predict how Roy would have fared without a mask.)

I honestly cannot wrap my head around the methodology. If we accept that hockey is in large part governed moreso by random events (at the game not individual level), then account for the fact that the individual variables are too great (x injured, who played with whom, who officiated, someone had a cold, the general pace of the game being different due to conditioning/training, the coaches use different methods, different rules, center lines/no center lines, and so on.....).... you cannot realistically compare anything across decades. The one argument I can accept to a certain extent, although this too may suffer from these same nefarious variables, is the relative standing amongst one's peers.

I am not against all attempts to do this because I think that there are broad strokes that can be painted and the process can stimulate conversation. I also have a personal bias that "math" can indeed be used to shed light.

There was a recent thread on the HoH Board comparing scoring streaks between eras.

http://hfboards.com/showthread.php?t=966625

I thought that the OP took a very balanced approach to the issue at hand. He/she accepted critisism of the method, accepted its limitations, tried several approaches, and refrained from any absolute conclussions. But the core point was made, hence I think the thread was a success.

 

[barneyg]

So under a google search for hockey Sabremetrics i came across the name Daryl Shilling. His site appears to no longer to be active and I couldn't find anything else about it. I guess my question to others on the board is, why has Sabremetric type analysis not taken off in hockey? wouldn't it make the debates we all have ranking prospects better to have concrete data to say "player x is better than player y"? Or, team a's roster is better than team b's roster based on the data? wouldnt it be helpful to know how much better a third liner on the Devils is than a replacement 4th line call up from the AHL?

I know hockey isnt baseball, its far less of a situational game and far more of a reactionary sport than baseball is, but I think there would be value in this kind of in depth statistical analysis, I even think that if the data was out there it could make scouting better and help teams make better personnel decisions. what say you hf boards?

As other posters have replied, there is some of this, although as usual you want to be careful with how you use the data. If you want a good place to start (and go back to regularly -- I do anyway) --

http://www.behindthenet.ca/

 

[Injektilo]

Perfect example:

Keith Jones was once (something ridiculous like) -7 in a game. And in the post-game review of the film, the coaches pointed out that he was on his man and where he was supposed to be on every single goal.

That's why you never look at a single game and make any sweeping statements.

The advanced stats are only useful over a long long time period, and they all have to be taken with other stats in mind. For example, players have no control over their goalie's SV% behind them, so any plus minus analysis needs to take that into consideration. A guy who is -20 might have just had crappy goaltending behind him, to no fault of his own. His goalie may have had a .905vsvp, but when he was on the ice it was .880.
And this example is also why corsi is better than straight plus/minus, bigger sample size and all.

 

[Sensfanman]

Perfect example:

Keith Jones was once (something ridiculous like) -7 in a game. And in the post-game review of the film, the coaches pointed out that he was on his man and where he was supposed to be on every single goal.

Well, firstly, +/- is junk so being a -7 doesn't mean anything. Secondly, it is possible the coaches were wrong.

 

[thinkwild]

I can sit in a stadium drinking a beer and relaxing and score a baseball game quite accurately, perhaps the occasional “is that an error or a hit” debate, but largely easy to measure events, easily attributable events, and easy to see how they help the team win events.

Its gotta be hard to keep a similarly accurate tally on all these hockey events. They are starting to gather a lot of data, but compiling it into meaningful information is still a challenge. But even for some of the easy stats, I still have a bit of confusion.

Like what is an assist? I thought I remembered as a kid that when someone scored off a deflection off the other teams goalie, there were no assists awarded.

I saw a play in Ottawa once where Kelly makes this great defensive decision to eat the puck in our zone as his teammates had abandoned him on a line change and the passing lanes were cut off. He took the hits, battled along the boards and finally managed to swing the puck off the glass where it bounced off the back of one our defenders ,who wasn’t looking, and into the open. Heatley raced to it, took off down the ice, wound up for a big slapper. It missed the net, the goalie steered it to his dman in the corner as it rebounded off the boards but it got lost in his dmans skates. Alfie came by and tried to swipe it between their dmans legs back to Spezza but it deflected off both the other teams dmens skates and into the net. A missed pass by alfie = goal. A missed net by heatley = assist. A didn’t notice the puck bouncing off his butt assist to our other dman. Kellys great play - unrewarded by any measure.

Or even what is a face off win? This one has always puzzled me. If 10 fans independently counted face off wins and then compared notes after the game, would any 2 agree? No ties, really?

Even hits. When is it a hit and not incidental contact. What if the hit causes a turnover vs. you make the hit and then fall down? If it’s a light nudge against the boards 2 seconds after the pass made to no consequence is it still a hit like one that changes the momentum of a game?

How is line between giveaway/takeaway determined? Home ice?

In tandem with developing and interpreting new stats, i could sure benefit from a review of some of the old simple ones anyway.
,