Randy Carlyle/Leaf Coach thread

[bustaheims]

So why not track that information and make it easily accessible? Zone starts and zone entries fall under these analytic stat categories. Is Nonis saying that he doesn't think it's valuable to know which players are more successful at entering the zone than others? Or if a players point totals are somewhat effected by how a coach is deploying him on face-offs?

No. What I think he's saying is that it's not worth hiring additional staff to track these things, nor is it necessarily important to give them a specific name or to measure them to a high degree of accuracy. If you observe players long enough with your pro and amateur scouts and make rough approximations of how they fare in these areas, you're going to come to the same basic conclusion as you would having additional staff tracking them and getting specific numbers on each player.

But nobody is suggesting that corsi should be looked at in a vacuum. And neither should +/-/. I think +/- can be a pretty useful stat if you take into consideration what type of ice-time a player is getting and who he's playing against. I'm sure many in the NHL community would agree with that. It's not perfect, but there is some value there. I would say the exact same thing about corsi.

The problem is that a lot of people do look at in a vacuum - or, at the very least, they're expressing conclusions based on Corsi without acknowledging the obvious flaw in the measurement while compounding the issue by using Corsi to come up with a bunch of other statistics. Like +/-, Corsi might have some value when looked at over multiple seasons and seeing if there's a trend, but, even then, as far as I'm concerned, the value is pretty limited.

And I think this is my biggest problem with how Nonis views corsi. It seems like he thinks the idea about corsi is that team with a high rating will win and a team with a low rating will lose. He said today that "we've had teams in the past where we were outshooting teams on a nightly basis. Our so-called Corsi stat was probably pretty good." It's like he's pretending that the team didn't have the worst goaltending in the league those years.

I don't think he's pretending that the goaltending was awful over those years. He was part of the management team that repeatedly looked to improve the situation, so, I'm sure he's well aware of the situation there. What's he doing is pointing out a flaw in how some people perceive and present Corsi and showing just how limited having a strong Corsi number can be.

Nonis and co. seem to be in the minority here though. In fact I honestly can't name a single other GM that has spoken out against advanced stats like he has. Maybe you can point out a few. But earlier this season somebody compiled a list of all the NHL teams that are using advanced stats or analytics one way or another. 17 teams seem to say that analytics are a valuable resource. That's including all the recent Stanley Cup champions. Winnipeg wasn't on that list but after what Cheveldayoff said today you can add him too.

I'm not sure I agree with you here. What I see in that article is a lot of teams using their own proprietary set of advanced stats and a small handful openly using things like Corsi. No one, including Nonis, is saying there's no value in finding advanced stats or analytics - just that they felt the companies that have presented to them don't provide enough to justify spending money on them. If they didn't feel there was any value in these things, why would they have met with a dozen different analytics people (according to Mirtle)? Clearly, they see a potential value in the area, just not in what's out there right now.

 

[CarltonTheBear]

So, yeah, I mean I get the problem with Nonis' comment if the idea is a war between fancy stats and reading the sports pages but I don't have any problem with Nonis looking at what exists in the world of advanced stats in Hockey nowadays, not thinking they outweigh or add much to the luxury of a professional scouting department and not devoting a lot of time/energy to them.

I was in the middle of writing a more detailed response but figured I would just sum up my thoughts a little more neatly. Basically, I do agree with some of your concerns regarding corsi. It's not a perfect stat and of course nobody is saying that it is. But I do think some of those concerns could just as easily be attributed to a stat like goals. You say that corsi isn't specific enough and that if you were a GM you would still want to scout a player to get actual specifics of what caused his corsi rating. Well, if I was a GM I would want to scout a player to see how he scores his goals too. Is he creating his own chances or being set up by a superior linemate? Is he scoring goals off the rush or from in front of the net? Is he scoring most of his goals late in blowouts or are they when the game is close? He is lighting up inferior opponents or playing against tough competition? Incidentally analytics could help answer a couple of those questions but I didn't even mean that. I just mean goals don't replace scouting either, but it doesn't mean they aren't valuable.

Also, just because you brought it up, it was mentioned earlier in this thread that corsi actually lines up very closely with time of possession/zone time stats. And really that's all corsi is trying to do, determine who controls the play longer. I think that's a valuable stat to have.

 

[bustaheims]

Also, just because you brought it up, it was mentioned earlier in this thread that corsi actually lines up very closely with time of possession/zone time stats. And really that's all corsi is trying to do, determine who controls the play longer. I think that's a valuable stat to have.

Sure, but I think that also supports Nonis and Co.'s position on bringing advanced stats and analytics people into the front office. Nothing that Corsi provides here can't be determined through traditional observational methods, either. It's really just another way of expressing things that people have been watching for since people starting scouting the game. The only thing it really provides over traditional scouting style observations is a number.

 

[Kin]

I was in the middle of writing a more detailed response but figured I would just sum up my thoughts a little more neatly. Basically, I do agree with some of your concerns regarding corsi. It's not a perfect stat and of course nobody is saying that it is. But I do think some of those concerns could just as easily be attributed to a stat like goals. You say that corsi isn't specific enough and that if you were a GM you would still want to scout a player to get actual specifics of what caused his corsi rating. Well, if I was a GM I would want to scout a player to see how he scores his goals too. Is he creating his own chances or being set up by a superior linemate? Is he scoring goals off the rush or from in front of the net? Is he scoring most of his goals late in blowouts or are they when the game is close? He is lighting up inferior opponents or playing against tough competition? Incidentally analytics could help answer a couple of those questions but I didn't even mean that. I just mean goals don't replace scouting either, but it doesn't mean they aren't valuable.

Sure, but on the other hand nobody would take issue if Nonis said something along the lines of "We don't judge players by goal totals, we prefer to actually see how they contribute offensively" or think that he was abandoning the future. I think the biases and influences on a player's goal totals are well known and well accepted.

The difference, and this more applies to the concept of CORSI as an individual measurement, is that I think there's still a specificity to goal scoring that corsi lacks. While a GM has a responsibility to account for how various playing styles will fit in to what he's putting together I think there's a widely held and relatively correct belief that, Cam Neely or Pavel Bure, one goal is as good as the other and "goals" is a fairly precise measurement of how good someone is at finding the back of the net. With Corsi it still doesn't seem like you're measuring any one thing and that someone can achieve a good rating by means of excellent defensive or offensive play, in which case I think it's still more useful to measure those things individually rather than glomming them all together under the heading of "puck possession".

Also, just because you brought it up, it was mentioned earlier in this thread that corsi actually lines up very closely with time of possession/zone time stats. And really that's all corsi is trying to do, determine who controls the play longer. I think that's a valuable stat to have.

Well, and excuse me if I'm misreading that, isn't that just one measurement of a handful of Maple Leaf games? And, even if true, doesn't that still make the argument for possession/zone time over Corsi?

 

[leafplasma]

Here we go!

LMAO!!!!

 

[bustaheims]

What I read here is the Nonis is a proponent of advanced stats in theory, but not where they're at right now - and I agree with him.

http://www.theglobeandmail.com/sports/hockey/leafs-beat/mirtle-nonis-not-yet-a-believer-in-hockeys-statistical-push/article15384253/?cmpid=rss1&click=dlvr.it

Nonis, however, remains hopeful that at some point there will be a numbers-based approach that the Leafs front office can get behind and use that isn?t currently available.

?People run with these stats like they?re something we should pay attention to and make decisions on, and as of right now, very few of them are worth anything to us,? he said at one point during the panel, blaming media and fans for overhyping the analytics currently available.

?That will change. I predict this [SportVU type of] technology is going to help us. There will be a stat that will help us make some decisions; I think that is coming.?

 

[OldTimeHockey]

Well, if I was a GM I would want to scout a player to see how he scores his goals too. Is he creating his own chances or being set up by a superior linemate? Is he scoring goals off the rush or from in front of the net? Is he scoring most of his goals late in blowouts or are they when the game is close? He is lighting up inferior opponents or playing against tough competition?

But is that not essentially what a Scout does now? Why bring in a guy to analyze those things you mentioned when your scouts are already paid to do just that.

 

[Frank E]

Are most of you are saying that Corsi doesn't necessarily help qualify performance in and of itself, although it may quantify a couple of measurements of performance?

 

[bustaheims]

Are most of you are saying that Corsi doesn't necessarily help qualify performance in and of itself, although it may quantify a couple of measurements of performance?

It attaches a quantity to a broad stroke of performance, but doesn't really help to provide details. And, really, that particular area of performance is one that good teams are constantly striving to improve any way, so, it's not really one that necessarily needed to be quantified for them.

 

[princedpw]

I mean, really, most of these stats are pretty basic and are often based on things that are either easily observable (like zone starts, for instance, can be largely be discovered through traditional scouting)

I don't think the advanced stats guys claim these things are "sophisticated".  In fact, these days, they often try to bend over backwards to say the opposite.

These things are certainly easily observable.  There may well be a difference though between systematically tracking these quantities vs. casually noticing gross trends from time to time.  The problem with the latter is that people just aren't that good integrating large volumes of data in their head.

It's a basic rule of science:  if you don't take down the data, do some experiments and test your hypotheses then you won't know what you are missing or what mistakes you are making. 

That's all I'd ask Nonis to do (as LK said):  Take down all the data you can and have some guys look at it.  Compare it to what your scouts give you.  See whose data is the better predictor, and in which circumstances.

... or suffer from similar flaws to existing stats when looking at individual players (Corsi/Fenwick are both just +/- for shot attempts, and, really, that means they suffer from the same basic issue as +/- itself - it's difficult to attribute much of anything to the individual player).

There's a pretty big flaw in the comparison between Corsi and +/-.  Yes, they are both similar in the sense that they are imperfect.  However, there are so many more shot attempts than goals that a lot of the noise/error in +/- is washed out of Corsi.  Again, that's not to say that Corsi is perfect but it is much, much better empirically.  So much better that a comparison to +/- isn't really helpful in understanding the state of affairs.  And no, Corsi on its own just isn't that much help.  One needs more context.

That is really not surprising though --- it's a general property of data analysis that more data can make an analysis vastly better.

 

[Kin]

There's a pretty big flaw in the comparison between Corsi and +/-.  Yes, they are both similar in the sense that they are imperfect.  However, there are so many more shot attempts than goals that a lot of the noise/error in +/- is washed out of Corsi.

I don't necessarily think that's true. While there are definitely more shots than goals, I think that there's a lot more randomness/outside influences that determine shots when a player is on the ice and I'd suspect that washes away a lot of the difference. I mean, people do forget that +/- does an ok job at identifying good players as well.

 

[Kin]

Are most of you are saying that Corsi doesn't necessarily help qualify performance in and of itself, although it may quantify a couple of measurements of performance?

In a team sense, sure. In an individual sense the problem is, as busta has said, similar to the problem with +/- where group results are being confused for individual performance.

 

[mr grieves]

Are most of you are saying that Corsi doesn't necessarily help qualify performance in and of itself, although it may quantify a couple of measurements of performance?

In a team sense, sure. In an individual sense the problem is, as busta has said, similar to the problem with +/- where group results are being confused for individual performance.

That's the problem with the bottom-line Corsi number, yeah. But what if you were to take it a game at a time and look what happens when line mates change? Couldn't one, say, look at the effect of swapping Kadri for Bozak on the top line and draw some conclusions about how the two group work together, what one player seems to contribute relative to another?

 

[Kin]

That's the problem with the bottom-line Corsi number, yeah. But what if you were to take it a game at a time and look what happens when line mates change? Couldn't one, say, look at the effect of swapping Kadri for Bozak on the top line and draw some conclusions about how the two group work together, what one player seems to contribute relative to another?

Sure, although it seems to me as though to really get a sense of that you'd have to account for the other two Leafs on the ice as well, but even then I don't know how much it would tell you. I mean, if you're the Leafs and you have the resources they do I'm not sure what the value is there vs. having someone watch every game specifically for that and grade every player on every shift. That's not to say you couldn't have both, and contrary to what Nonis might say I'm sure there is someone within the organization who's broadly aware of those sorts of things, just that I don't know what Corsi would tell you that the other thing wouldn't.

 

[moon111]

Politicians use stats all the time.  Enough said.

 

[GhostofPotvin29]

Worth a read: http://www.mc79hockey.com/?p=6435

 

[princedpw]

What I read here is the Nonis is a proponent of advanced stats in theory, but not where they're at right now - and I agree with him.

http://www.theglobeandmail.com/sports/hockey/leafs-beat/mirtle-nonis-not-yet-a-believer-in-hockeys-statistical-push/article15384253/?cmpid=rss1&click=dlvr.it

Nonis, however, remains hopeful that at some point there will be a numbers-based approach that the Leafs front office can get behind and use that isn?t currently available.

?People run with these stats like they?re something we should pay attention to and make decisions on, and as of right now, very few of them are worth anything to us,? he said at one point during the panel, blaming media and fans for overhyping the analytics currently available.

?That will change. I predict this [SportVU type of] technology is going to help us. There will be a stat that will help us make some decisions; I think that is coming.?

This is great if he believes this.  What I'd like to see though is the leafs attempting to get ahead of the game.  Instead of passively waiting for something to come along, and when it does, jump on the bandwagon like everyone else,  I'd like them to get ahead of the game.

 

[princedpw]

Politicians use stats all the time.  Enough said.

Yes, stats can be very misleading.  However, your intent and my intent, as fans, would not be to intentionally mislead, it would be to uncover more of the truth behind who is playing well and who isn't, who we think should stay on the team and sho shouldn't.  We have a different set of incentives than your average politician...

 

[CarltonTheBear]

http://www.theglobeandmail.com/sports/hockey/leafs-beat/mirtle-nonis-not-yet-a-believer-in-hockeys-statistical-push/article15384253/?cmpid=rss1&click=dlvr.it

?That will change. I predict this [SportVU type of] technology is going to help us. There will be a stat that will help us make some decisions; I think that is coming.?

If the Leafs aren't one of the first teams to get those cameras installed use the cameras I'll be pretty disappointed. Busta I'm guessing you know a little more about them than I do since you follow basketball, is there a reason they aren't being used in the NHL right now?

edit: Just realized that the cameras themselves would already be installed in the ACC for the Raptors. I wonder if they would need brand new ones for hockey or if there just needs to be some tinkering done.

 

[nutman]

So you guys need stats for what? if I watch a player enough and keep tabs on his basic stats (the ones we have been using for years and years) then getting a picture of how good he is comes very easy. All this corsi and other detailed stats is nothing but overkill. Remember it is a game, and dissecting players that far from game to game is just insane as a fan. As for management they have there ways to judge players, and most don't use all these new made up stats. I guess the only ones who do need these stats are the people who cant judge how good a player is.
JMO...