Confirmed with Link: Canes Sign De Haan To A Real Life Contract

[HisIceness]

It’s weird, at the beginning of free agency I saw one of those analytic guys say (and post the reference) that CDH was largely the same player, still. I think it was JFreshHockey on Twitter.

 

[Sens1Canes2]

Was I right? I don’t know how to read that stuff.

 

[HisIceness]

Was I right? I don’t know how to read that stuff.

Probably. De Haan himself asked if this meant he sucked and JFresh said its the opposite. I don't know how to read it either but JFresh has never lied to me before

 

[Discipline Daddy]

Probably. De Haan himself asked if this meant he sucked and JFresh said its the opposite. I don't know how to read it either but JFresh has never lied to me before

Not that anyone asked but I don't put much stock into JFresh's models. I think they're less accurate than Dom's models, and I don't think Dom's models are great either. To be fair, it's just really hard to make an algorithm to predict sports like hockey. I feel like a survey of hfboards fans is generally more accurate, but that is obviously also flawed and a lagging indicator. It's just hard to do. I don't think the eye test will ever go away.

 

[AhosDatsyukian]

Some team should GM by HF Boards polling on potential moves and see how it works out.

 

[Discipline Daddy]

Some team should GM by HF Boards polling on potential moves and see how it works out.

LOL that would be such a fun disaster. My point is that a human element, while flawed, is important to include in a model. I don't see how you can scout without having scouts, for example.

 

[Boom Boom Apathy]

LOL that would be such a fun disaster. My point is that a human element, while flawed, is important to include in a model. I don't see how you can scout without having scouts, for example.

I agree with your comments on the difficulty of models. The problem with the human element is that no human can see enough games for enough players to be an adequate sample size. Scouts are paid to do so and even they can't; and they are more often wrong than right. We fans and even media have virtually no chance at seeing enough actual games for most players.

I probably saw de Hann play 4-5 times on TV last year. Really not enough to make an overly meaningful assessment, particularly on TV where the cameras only follow the puck and not every situation is encountered. I think what most try to do is pick up trends or tendencies that players have and not really get a great overall assessment. Do we see ill timed pinches? Does a player tend to stay on the perimeter? Does a he panic when pressured? Does his skating look sub par? etc..

Obviously, we watch our own team enough to get a better overall feel, I'm only referring to other teams.

The 2nd problem with the human element is that it's impossible to remove our own biases. If we are a person that values strong defensive play, or physical play vs. a person that values more offensive creativity and skill, we'll naturally included our biases no matter how hard we try not to.

I get that the models are trying to do what the eye test can't, remove bias and use a bunch of data to get an overall assessment of a player, but as you said, when you remove the human element, it leads to other issues with the model.

 

[MinJaBen]

Some team The Leafs should GM by HF Boards polling on potential moves and see how it works out.

This would be the most fun.

 

[cptjeff]

Was I right? I don’t know how to read that stuff.

It's percentiles among peers. So his WAR (Wins Above Replacement, a voodoo number intended to be an overall player score) is better than 65% of NHL defensemen, his EV defense is better than 74%, but his EV offense is only better than 34%, which means 64% of defenseman in the NHL are better than him at creating points at even strength. He makes his teammates better, and is better than 95% of other D in that regard. Plays against good competition, doesn't take many penalties. The only weak spot we care about (he ain't here for offense) is that his penalty killing is, by these metrics, not good. But that could be a data artifact based on playing in Chicago (as could some of the more positive numbers, because these are still shot attempt based stats, which are not terribly reliable in general). But broadly it means he's been an effective defensive D that's probably a solidly middle pairing guy on average across the league, and that makes him quite the bargain at 800k and a great get for the 3rd pair.

 

[LostInaLostWorld]

The 2nd problem with the human element is that it's impossible to remove our own biases. If we are a person that values strong defensive play, or physical play vs. a person that values more offensive creativity and skill, we'll naturally included our biases no matter how hard we try not to.

I get that the models are trying to do what the eye test can't, remove bias and use a bunch of data to get an overall assessment of a player, but as you said, when you remove the human element, it leads to other issues with the model.

Agree. And models have biases as well.

 

[Derailed75]

The problem all sports face when it comes to prospects is that more than half the guys they are playing against when you scout them aren't getting anywhere near their collective big leagues. Its much easier to look good even great against guys that will never sniff the bigs. In the top leagues it takes a rare combination of skill, ability, determination, and practice. 10 or so years ago the Browns drafted a DB that had it all and not just at the combine when you watched the kid against top colleges he excelled. When he was put on the field in the NFL just didn't want it enough to put in the work Monday-Saturday and he looked like a clown on Sundays. Same thing applies to the NHL and all other leagues.

Its really hard to project "want to"!