Is Tulsky part of the problem?

[Navin R Slavin]

This is what, the third season where we're way at the top of the Corsi heap, right?

We hear, over and over, that "we're getting our chances" -- but our PDO continues to be world class terrible.

Are we optimizing for the wrong metrics? Are we confusing correlation with causation? Are we overvaluing players who drive shot numbers (Hamilton) and undervaluing players who can put the puck in the net?

Is the organization's reliance on fancy stats actively making us worse?

Click to read more...

 

[the halleJOKEL]

all signs point to maybe

 

[SlavinAway]

100%, unequivocally, definitely, maybe

 

[Tryamw]

100%, unequivocally, definitely, maybe

Except if it is not then definitely probably maybe. Or do i need more hedge.

 

[Anton Dubinchuk]

Are we optimizing for the wrong metrics? Are we confusing correlation with causation? Are we overvaluing players who drive shot numbers (Hamilton) and undervaluing players who can put the puck in the net?

Yes

Yes

Yes but bad example because until he got here he scored tons of goals.

I think it’s a combo. There’s also evidence that we are ruining guys with our play.

I think the Canes are actually proving to the world and to the analytics community that the “causation” between Corsi and success is weaker than we think (I’m sure I made any formally trained statisticians shutter with my verbiage there, but whatever). Clearly there’s correlation, and some of that correlation includes causation. In a vacuum, no one would argue that outshooting the opponent is a good thing. But we’ve made it such an emphasis that we’ve lost the actual way to win hockey games (outscoring, not outshooting).

The extreme to which we’ve taken this is similar to a team spending $30 million per year of the cap on a great goalie because “there’s a strong correlation between good goaltending and winning.” Yes, there is. But going to the extreme of spending that portion of your cap on it isn’t the key to success. It’s the same thing we’ve done with Corsi.


Honestly, I think the further we separate ourselves from the pack from a Corsi +/- perspective, it doesn’t make us “unluckier” and more likely to turn it around. I think it simply provides more evidence for Hank’s point. We’ve stopped caring about winning games, thinking that the causation is “stronger” than it is.

 

[SlavinAway]

Dellow: The next generation of data will drastically change...

Whether it's a problem for us or not I don't think fancy stats are going to go away and in fact will probably get fancier...

"Hockey doesn’t really work like that. The sport doesn’t lend itself to easily tracking the things that result in goals being scored or prevented. Historically, hockey leagues have tracked goals and assists. While that’s useful information, it’s not unlike runs and RBI in baseball: an attempt to hand out credit after the fact rather than tracking the building blocks of goals. Shot attempt data and expected goals models are helpful but there’s a huge issue with a lack of information about how the puck moved and where the non-shooting players were when the puck was shot. That’s the information that’s analogous to on-base percentage and slugging percentage in baseball.

You can think of hockey as being two cycles of analysis behind baseball. We don’t really have the equivalent of OBP and SLG (certainly not to the same degree of usefulness) and we definitely don’t have the kind of process data that MLB teams now enjoy, with every player’s movements being tracked. If the NHL gets player and puck tracking right, these two levels of analysis will suddenly become possible.

What a system like this will generate is a file telling you where every player on the ice was located at various points in a game. If it recorded player/puck location 20 times per second and it was one of those increasingly frequent Maple Leafs games in which nobody gets a power play, you’d end up with 20 (data points per second) * 3600 (seconds) * 13 (players and puck) = 936,000 lines of information about a game. The current play-by-play files are usually around 300 lines, so that’s a bit of a difference.

Away from the ice, one of the real challenges of hockey is allocating credit or blame between players. This is particularly true when dealing with players who play with superstars – every partner Nicklas Lidstrom ever had posted great numbers – or players who are playing on particularly good or bad teams. The ability to better isolate what players are contributing away from their linemates will result in much better evaluations of players who are in unusual circumstances.

In order for all of this to take place, there’s going to be some changes in how front offices and coaching staffs operate. There’s a saying that you hear around hockey: “Managers manage, coaches coach and players play.” This doesn’t really make a ton of sense, even without really good data. If the manager builds a team a certain way, how does it make sense to have a coach who may want to run it another way? If a coach or player has insights that can be used by the front office in modelling how the game works, why not draw on them? One of the themes running through Big Data Baseball, as well as in any discussion of successful Major League Baseball teams these days, is how collaborative the front office, coaching staff and analytics departments have become. Hockey’s historic model of a general manager who builds a team and hands the keys over to the coach will lose out to teams that adapt a more collaborative approach in order to maximize the value of the new information that’s available to them."

 

[Navin R Slavin]

Analytics will continue to improve, yes. And that's fine.

But effective use of analytics must ultimately lead to strategies that lead to wins.

It seems that at some point the front office would say "ok so maybe Corsi isn't as highly correlated to winning as we thought, so maybe we should drive other metrics that might."

Instead, it seems as though they're doing the opposite: they're driving for historic epic Corsi. And the results on the scoreboard are no better.

 

[TheBigKahuna]

100%, unequivocally, definitely, maybe

You'll need stats to back up your point.

 

[SlavinAway]

Analytics will continue to improve, yes. And that's fine.

But effective use of analytics must ultimately lead to strategies that lead to wins.

It seems that at some point the front office would say "ok so maybe Corsi isn't as highly correlated to winning as we thought, so maybe we should drive other metrics that might."

Instead, it seems as though they're doing the opposite: they're driving for historic epic Corsi. And the results on the scoreboard are no better.

Maybe they have found a new metric they're working to optimize that extremely high Corsi is just a side effect of.

Honestly we'll probably never know from the outside; all of the Corsi talk isn't coming from inside the organization.

 

[MinJaBen]

It seems that at some point the front office would say "ok so maybe Corsi isn't as highly correlated to winning as we thought, so maybe we should drive other metrics that might."

Who says that Tulsky hasn't said this but we are unable to effectively make the changes? At this point, I'm starting to wonder if Rod knows how to change his socks he's so fixated on keeping things the same.

 

[Navin R Slavin]

Who says that Tulsky hasn't said this but we are unable to effectively make the changes? At this point, I'm starting to wonder if Rod knows how to change his socks he's so fixated on keeping things the same.

We can't know, of course. But if Tulsky is making these points and the org isn't responsive, that's a whole separate issue, which would imply that Dundon's collaborative front office model isn't working.

 

[Hulkacaniac]

Analytics will continue to improve, yes. And that's fine.

But effective use of analytics must ultimately lead to strategies that lead to wins.

It seems that at some point the front office would say "ok so maybe Corsi isn't as highly correlated to winning as we thought, so maybe we should drive other metrics that might."

Instead, it seems as though they're doing the opposite: they're driving for historic epic Corsi. And the results on the scoreboard are no better.

Kind of reminds me of those clickbait articles like “seven habits of highly successful people” as if doing this one thing will turn you into a millionaire.

 

[AD Skinner]

Obviously I'm not privy to any of these conversations but I kinda wonder how many are really taking place. The front office hasn't made a move since the season started other than McElhinney and moving players to/from Charlotte. Maybe the moves aren't there to be made but several teams have made change of scenery type trades. I think we are fast approaching where a "trade for trading's sake" would be justified.

 

[Bub]

My addled brain read this thread as "Is Tlusty part of the problem?" and I thought "Wow, we really ARE feeding a fed horse here aren't we."

 

[MinJaBen]

I think we are fast approaching where a "trade for trading's sake" would be justified.

Frankly, with the group that we have making the decisions, this scares me.

No way they don't somehow screw it up.

 

[My Special Purpose]

If creating chances isn't the issue, then it's got to be converting chances, right?

I think what we're finding is that all NHL players aren't the same. Why we didn't know this already is beyond me, yet here we are. A grade-A chance generated for Phil DiGiuseppe does not equal a grade-A chance created for Timo Meier. A grade-A chance created for Jordan Staal does not equal a grade-A chance for Joe Pavelski, etc.

For example, let's say we have 20 scoring chances in a game and our opponent has 15. Our 20 chances break down like this: McGinn 4, Wallmark 3, Staal 3, Martinook 2, Aho 2, Teravainen 2, Foegele 1, Faulk 1, DiGuiseppe 1, Svechnikov 1. Our offensive style is set up to create chances for guys who can't score. But of course, all we have is players who can't score. So, it's not the style, it's the players? Yeah, we need better goal scorers. But what we can do is a better job of putting players with better scoring ability (Svechnikov, Teravainen, Ferland when he's healthy) in the positions that are currently occupied by McGinn, Staal, etc., in an effort to convert more chances, but in the end, we have very limited options currently under contract.

Meanwhile, our opponents' break down like this: Meier 3, Hertl 3, Pavelski 2, Burns 2, Karlsson 2, Kane 1, Sorensen 1, Goodrow 1.

Who do you think is going to convert more of their chances?

We've tried to "Moneyball" the system by not paying top $$ for goal-scorers, thinking that guys who possess the puck and generate chances, but who struggle to score, are undervalued. The thinking is that the goals will be a natural result of having several of these players on the same team, dominating possession and shot generation. We're (hopefully) learning that it doesn't work and that shooting percentage is a skill, and not a product of luck.

 

[MinJaBen]

For example, let's say we have 20 scoring chances in a game and our opponent has 15. Our 20 chances break down like this: McGinn 4, Wallmark 3, Staal 3, Martinook 2, Aho 2, Teravainen 2, Foegele 1, Faulk 1, DiGuiseppe 1, Svechnikov 1.

I know these are made up numbers, but it does illustrate a point. It is the GMs job to change the players on this list. It is the coaches job to change the order of them on this list. If our list does actually look like this, then both of them are not doing their jobs well.

 

[SvechneJerk]

If creating chances isn't the issue, then it's got to be converting chances, right?

I think what we're finding is that all NHL players aren't the same. Why we didn't know this already is beyond me, yet here we are. A grade-A chance generated for Phil DiGiuseppe does not equal a grade-A chance created for Timo Meier. A grade-A chance created for Jordan Staal does not equal a grade-A chance for Joe Pavelski, etc.

For example, let's say we have 20 scoring chances in a game and our opponent has 15. Our 20 chances break down like this: McGinn 4, Wallmark 3, Staal 3, Martinook 2, Aho 2, Teravainen 2, Foegele 1, Faulk 1, DiGuiseppe 1, Svechnikov 1. Our offensive style is set up to create chances for guys who can't score. But of course, all we have is players who can't score. So, it's not the style, it's the players? Yeah, we need better goal scorers. We can do a better job of putting players with better scoring ability (Svechnikov, Teravainen, Ferland when he's healthy) in the positions that are currently occupied by McGinn, Staal, etc., in an effort to convert more chances, but in the end, we have very limited options currently under contract.

Meanwhile, our opponents' break down like this: Meier 3, Hertl 3, Pavelski 2, Burns 2, Karlsson 2, Kane 1, Sorensen 1, Goodrow 1.

Who do you think is going to convert more of their chances?

We've tried to "Moneyball" the system by not paying top $$ for goal-scorers, thinking that guys who possess the puck and generate chances, but who struggle to score, are undervalued. The thinking is that the goals will be a natural result of having several of these players on the same team, dominating possession and shot generation. We're (hopefully) learning that it doesn't work.

This. All day.

Good post, Kev.

 

[bleedgreen]

I think the money ball fancy stat approach only works for your bottom six when you’re trying to squeeze some points out of a huge selection of bottom six available guys. Your top six forwards should be guys that are eye test worthy offensively, as in I can clearly tell that guy can score.

Because I’m watching him score.

 

[Finlandia WOAT]

It seems that at some point the front office would say "ok so maybe Corsi isn't as highly correlated to winning as we thought, so maybe we should drive other metrics that might."

Instead, it seems as though they're doing the opposite: they're driving for historic epic Corsi. And the results on the scoreboard are no better.

They're not trying to game Corsi. Team employed data guys have their own models and stats which are private (their large salaries are in part to keep it that way).

An argument can be made that under Peters they were trying to game shot distance without risky cross ice passes- which is why they took a ton of point blank shots to the goalies' left and a ton of shots from the halfwall/point.

I do think Tulsky was a voice in moving Skinner and acquiring Hamilton.

Imo their good corsi numbers are a function of, in no order: lack of talent up front, esp at center; lack of cross ice passes that force goalies to move laterally, done in part because of the lack of talent and in part to prevent interceptions and odd man rushes (peters did this too); firing long distance bad angle shots in lieu of cross ice passes in hopes for a rebound, pinball or fluke goal; centers who get boxed out/moved easily in the slot; a defense that is very good at defending the rush and defensive zone possession; playing from behind a lot, which influences shot totals (this was the main reason their peripherals were so good in 17-18, imo), and a League wide shift away from a style of play for which corsi served as a useful indicator, which benefits periphery teams like the Hurricanes.

 

[Big Daddy Cane]

8 of the 10 best offensive clubs in the NHL have at least one former Top 3 pick in their forward group. Those include Tampa (Stamkos), Colorado (MacKinnon and Landeskog), Toronto (Matthews and Tavares), Washington (Ovechkin), Ottawa (Duchene), Winnipeg (Laine), Pittsburgh (Crosby and Malkin) and Florida (Barkov and Huberdeau.) Calgary and Columbus are the two that built without the benefit . Even in Columbus’ case, Panarin is the fruit of the Rick Nash trade tree.

This franchise has been handcuffed for a while because they never committed to rebuilding and/or never got lucky with the lottery until this year. We just have to be a little patient here. It's not going to be this year, obviously. Svechnikov’s development will cure a lot of what ails this franchise. Aho, Svechnikov, Teravainen and Necas has the makings of a legit forward group. I'd like to see one more core piece added to that group long-term, but it can be the 4th or 5th best player among those.

 

[A Star is Burns]

I'd say anyone that's been associated with the franchise for a few years or more is part of the problem. I think in Tulky's case, we don't have the slightest clue how much what he does impacts us, but I'm sure it does have some impact on our woes.

 

[CandyCanes]

8 of the 10 best offensive clubs in the NHL have at least one former Top 3 pick in their forward group. Those include Tampa (Stamkos), Colorado (MacKinnon and Landeskog), Toronto (Matthews and Tavares), Washington (Ovechkin), Ottawa (Duchene), Winnipeg (Laine), Pittsburgh (Crosby and Malkin) and Florida (Barkov and Huberdeau.) Calgary and Columbus are the two that built without the benefit . Even in Columbus’ case, Panarin is the fruit of the Rick Nash trade tree.

This franchise has been handcuffed for a while because they never committed to rebuilding and/or never got lucky with the lottery until this year. We just have to be a little patient here. It's not going to be this year, obviously. Svechnikov’s development will cure a lot of what ails this franchise. Aho, Svechnikov, Teravainen and Necas has the makings of a legit forward group. I'd like to see one more core piece added to that group long-term, but it can be the 4th or 5th best player among those.

Almost right. But Dubois was a 3rd overall pick for the Blue Jackets

 

[sheriff bart]

I think the problem is the 300 level hot dog guy.

Seriously, I think everybody in the office is carrying some responsibility. If Tulsky's fancy numbers aren't getting it done on a budget, then maybe it can't be done on a budget. I think good NHL teams are all about identifying, drafting, and developing talent more than advanced stats

 

[Blueline Bomber]

I'd say anyone that's been associated with the franchise for a few years or more is part of the problem. I think in Tulky's case, we don't have the slightest clue how much what he does impacts us, but I'm sure it does have some impact on our woes.

I mean, yes and no. Obviously, there's a lot of blame to go around, but it's interesting that so many people that failed hard while with this organization have found more success outside of it. Maurice, Peters, Staal, Lindholm, Skinner, etc.

Would it really surprise anyone if we let Tulsky go, or Brindamour, or even someone like Rask or Wallmark and they suddenly become valued pieces that lead to the success of whatever franchise we lose them to?