The Advanced Stats Thread Episode VIII:

[Machinehead]

Ok so maybe Carrick isn't as great as the possession metrics indicate.

Do I still take him over 99 billion percent of the garbage on NHL third pairs? You better know it.

 

[silverfish]

I would've been able to calculate the threshold ixFsh% where a shot attempt turns from a detriment to a positive, but my coefficient CSV that R printed for me is broken and any coefficient that starts with anything other than "0," is just a long string of numbers without a decimal point.

Going by which decimal placement produces reasonable results, a shot from a defenceman must have an ixFSh% of at least 1.067% to contribute positively to a projected xGF%Rel. I.E. as long as it's not blocked it's basically good, but the higher the quality the better.

An example of how it comes out; Brent Burns' 22.17 iCF/60 and 0.38 ixGF/60 gives his projected xGF%Rel a +2.55 while Anthony DeAngelo's 11.41 iCF/60 and 0.3 ixGF/60 gives him a +3.17. However, the model has no idea that DeAngelo has an airsoft while Burns has a 7.62 cal SIG516.

What's the strength of your projection model?

It seems very odd that you get "penalized" for having a higher iCF60. I suppose intuitively, it makes sense, but every iCF contributes to ixGF, regardless of how small the contribution is.

 

[Blue Blooded]

What's the strength of your projection model?

My CSVs are all f***ed and I don't have R installed on this laptop yet, so unfortunately I can't say exactly right now.

A much more rudimentary version of the model (pre-xGF, April 2015) would predict the CD/60 of a defenceman moving to a new team at 0.1833 and predict the metric itself at 0.3579. For reference, CD/60RelTM would in the same scenario predict CD/60 at 0.0850 and itself at 0.1258.

 

[Blue Blooded]

It seems very odd that you get "penalized" for having a higher iCF60. I suppose intuitively, it makes sense, but every iCF contributes to ixGF, regardless of how small the contribution is.

IIRC only unblocked attempts add to ixGF, blocked shots are a 0 since they don't have coordinates. That's why you have ixFSh% and not ixCSh%.

 

[Blue Blooded]

IIRC only unblocked attempts add to ixGF, blocked shots are a 0 since they don't have coordinates. That's why you have ixFSh% and not ixCSh%.

This makes me want to try the model with iFF/60 separated from iBlockedShots/60 to see if it improves. Though I may already have tried this and forgotten about it, I know there are many of approaches I've tried 2+ times because I forgot that I already tried it since I don't write this stuff down enough.

 

[silverfish]

My CSVs are all ****ed and I don't have R installed on this laptop yet, so unfortunately I can't say exactly right now.

A much more rudimentary version of the model (pre-xGF, April 2015) would predict the CD/60 of a defenceman moving to a new team at 0.1833 and predict the metric itself at 0.3579. For reference, CD/60RelTM would in the same scenario predict CD/60 at 0.0850 and itself at 0.1258.

Gotcha.

IIRC only unblocked attempts add to ixGF, blocked shots are a 0 since they don't have coordinates. That's why you have ixFSh% and not ixCSh%.

Right, my bad, all fenwicks since there's not attempt location data on blocked shots.

 

[Machinehead]

Sam Gagner is yet another unspectacular but very useful player who is now a pariah because he doesn't fit the bottom six paradigm.

 

[Filthy Dangles]

Sam Gagner is yet another unspectacular but very useful player who is now a pariah because he doesn't fit the bottom six paradigm.

Sam Gagner is probably closer to 'serviceable' than a 'very useful' player. Don't forget he comes with a 2 year and 6M dollar commitment....I really don't see what makes him valuable at this point.

 

[Machinehead]

Sam Gagner is probably closer to 'serviceable' than a 'very useful' player. Don't forget he comes with a 2 year and 6M dollar commitment....I really don't see what makes him valuable at this point.

He isn't actively bad. That makes him valuable. I'd wager there's an actively bad player in most if not all opening lineups. Maybe two.

As we've seen many times over, this league prefers terrible players to meh players.

 

[Filthy Dangles]

He isn't actively bad. That makes him valuable. I'd wager there's an actively bad player in most if not all opening lineups. Maybe two.

As we've seen many times over, this league prefers terrible players to meh players.

He's overpaid, that's all. If he cost 1/3 of what he did there would be takers. You can't be "not actively bad" and also be overpaid and expect people to be lined up for your services

 

[Machinehead]

He's overpaid, that's all. If he cost 1/3 of what he did there would be takers. You can't be "not actively bad" and also be overpaid and expect people to be lined up for your services

Fair point, in theory.

In practice, Jay Beagle and Brandon Sutter are f***ing horrendous and look what the Canucks pay them.

 

[GeorgeKaplan]

Seems like the Canucks are going to rely heavily on their top line and (most likely) trapping their way to a playoff berth (which they’re most likely not going to get)

 

[Machinehead]

Seems like the Canucks are going to rely heavily on their top line and (most likely) trapping their way to a playoff berth (which they’re most likely not going to get)

Jay Beagle's team has never shot the puck with him on the ice in his entire life, so he'll help the trap for sure.

 

[Machinehead]

Interesting data on Vesey

I do have to admit that the Rangers are getting great chances when they have the puck with Vesey on the ice. Problem is they have the puck like once a month when Vesey is on the ice.

 

[GeorgeKaplan]

Jay Beagle's team has never shot the puck with him on the ice in his entire life, so he'll help the trap for sure.

A hot take of mine that I was pretty quiet about for a long time was that JT Miller was essentially our Sam Gagner, I never looked into their shot metrics or anything, but they’re both just guys who rack up points on lower lines while not really being all that great to me

 

[Machinehead]

A hot take of mine that I was pretty quiet about for a long time was that JT Miller was essentially our Sam Gagner, I never looked into their shot metrics or anything, but they’re both just guys who rack up points on lower lines while not really being all that great to me

True, but I would take JTM over most bottom six guys (maybe not at his contract)

 

[Blue Blooded]

Spooner and Chytil on the same line against the Preds, what's the over/under CF%? 30?

 

[silverfish]

Spooner and Chytil on the same line against the Preds, what's the over/under CF%? 30?

I'm taking the over because score effects.

 

[Blue Blooded]

I'm taking the over because score effects.

I like hate how you had to qualify it with score effects.

 

[RGY]

Whats a good site to see how much ice time one player had with each individual teammate? And maybe relative to how they scored?

Looking into how much Tavares may have effected Barzal last year

 

[DanielBrassard]

Whats a good site to see how much ice time one player had with each individual teammate? And maybe relative to how they scored?

Looking into how much Tavares may have effected Barzal last year

Naturalstattrick is the best site now for that.

 

[aufheben]

Whats a good site to see how much ice time one player had with each individual teammate? And maybe relative to how they scored?

Looking into how much Tavares may have effected Barzal last year

Might try hockeyviz.com for visuals.

 

[aufheben]

You can look into Lee too who the Islanders just named their captain.

 

[silverfish]

Only saw the Preds empty net and Buch's goal. Charts fair or unfair vs eyes?

I'm taking the over because score effects.

RIGHT AGAIN!

 

[DanielBrassard]

I'd switch Namestnikov and Spooner.