The Advanced Stats Thread Episode V: Rick Nash Camera Stares/60

[Thirty One]

I keep forgetting to post this. But for the stat nerds there's a good book that came out a few months back written by Michael Lewis of Moneyball, The Big Short, and The Blind Side fame. It's about two psychologists that did research on human judgment and decision making and basically came to the conclusion that people make a lot of systematic mistakes. They make mental shortcuts that don't always correspond with reality. The book is called "The Undoing Project".

The Undoing Project

Stats Nerds 1- Eyeball test Proponents 0

I've read it. I actually referenced Tversky and Kahneman on Rangers Unlimited back in the day.

That's why I was so offended when Lewis wrote that he had never heard of them.

 

[Ola]

I'll leave this open for someone else to answer if they feel like it since i'm not good enough at digging up numbers and explaining how/why etc.

I think you are just too humble!

 

[silverfish]

Totally, man

 

[TheTakedown]

Totally, man

Sounds like a corporate BS answer.

Can't wait to watch the **** show that the king's become!

 

[Ola]

Sounds like a corporate BS answer.

Can't wait to watch the **** show that the king's become!

The key is how Rob Blake connects the metrics to their system. This is were it becomes very valuable for a team, ie by creating measure points.

Like look at it from the perspective of the work done in regards by Bengt Holmström of the Massachusetts Institute of Technology (MIT) and Oliver Hart of Harvard University which won them the Nobel Prize in Economics 2016. Especially Holmström's Informativeness Principle can be used directly on metrics in hockey.

In order for a principal (a team/or company with an employee) to change the behavior of the agent (the player/the employee), it works best to set up goals that (i) can be measured and (ii) give the principal information on how the agent is doing.

For example, if AV orders Brendan Smith to stop taking those effin stupid penalties, he should be able to expect reasonable results. It is relatively easy to measure the amount of stupid penalties Smith takes. When I played, we had no video. The coach before a game preached stuff like "we must be more focused with the puck, turn the play faster, be more alert, not make mistakes on the offensive blueline" and so forth lol, then in the first intermission or after a game he agitatedly screamed stuff like "why aren't you turning the play faster? why aren't you more focused?" and so forth. My point is just, these cliches that we all have heard are not very tangible. They are in principle pretty worthless measure points in the context of the informativeness principle. And this is exactly what you work at in principle when team uses videos. They take the coach words, point at the TV and makes it more tangible. Its a step forward.

But look at like the picture from Babcock's whiteboard in the TML's dressing room. They are clearly taking it another step further using metrics. Puck retrivals are an issue for TML. To go back deep into your own end, fetch a puck, successfully move it out of your own end. This is something they want to do better. Babcock have undoubtedly shown they video of how it should be done. How the 5 man unit must act. They are practicing on it. But how are metrics used to improve their work here? Its easy, they put a guy in the stands that puts numbers on what they are doing on the ice. Its as simple as that:
-Brendan Smith, I told you to stop taking stupid penalties, you have 24 PIMs the last 8 games, its not good enough.
-Morgan Reilly, you must do a better job retrieving pucks in your own end. You only did successfully 8 out of 14 times against Boston. It is not good enough, you must get your successful puck retrivals up towards 80-90%. Lidström could do it at 95%, its not impossible.

What is Rob Blake saying? A team have a system, by creating measure points for your team and players in relation to what you want to achieve on the ice, you will benefit from it and get better results. Its as simple as that.

Additionally, Holmström developed and published the “informativeness principle” in 1979, which addresses the “principal-agent problem” (the structure of contracts between employers and employees),” as MIT explains: This principle suggests that optimal contracts should structure compensation based on all outcomes that can potentially provide information about actions that have been taken.

http://bigthink.com/laurie-vazquez/what-contract-theory-is-and-why-it-deserves-a-nobel-prize

 

[sbjnyc]

People are proving that SVA CF% is a very good predictive stat, that's it.

Predictive stats need predictions don't they? Is this more predictive than regular season points or wins or goal differential? Cup finalists and effort finalists aren't anything special in this picture and there are no comparison of teams that miss the playoffs. This is just an eye test for corsi.

 

[Zil]

Where do I find game by game corsi numbers for the playoffs? Corsica isn't showing anything.

 

[Ola]

Where do I find game by game corsi numbers for the playoffs? Corsica isn't showing anything.

These numbers are of course very influenced by Zone starts and match-ups, but some lines had good games.

 

[Ola]

Predictive stats need predictions don't they? Is this more predictive than regular season points or wins or goal differential? Cup finalists and effort finalists aren't anything special in this picture and there are no comparison of teams that miss the playoffs. This is just an eye test for corsi.

I think Corsi had a very high predictive rate for a while, but the game has changed a bit the last year or two and the sample sizes aren't very big and will be impacted by a lot of factors.

I just react to the notion that its still kicking the alternatives, because Corsi has been all over the place. As long as the league looked like it did last year, I am also very skeptical about PDO.

But it might be that a big reason for this is because the group lead by Sutter, Claude Julien and their copy cats didn't adopt enough to the more high risk rush play-trend that has been so effective offensively in the NHL just the last little while. Hence there is an argument to be made that Corsi impact on future success was biased since a few leading teams had flawed systems, and if a bigger portion of the teams went in the same direction, Corsi would regain its position. I know many disagree, but I definitely think its a fact that the way Bos, LAK and co played this last season boosted corsi at the cost of effectiveness offensively, for reasons I have mentioned several times in length.

Goal diff the last 4 years beat Corsi by "two overall positions" roughly, adjust the above inconsistencies and Corsi might regain the lead.

 

[Ola]

Is there data out there on de facto match-ups?

 

[silverfish]

Where do I find game by game corsi numbers for the playoffs? Corsica isn't showing anything.

Natural Stat Trick is still posting game-by-game summaries: http://naturalstattrick.com/game.php?season=20162017&game=30111

 

[Mac n Gs]

Why doesn't Manny scrap playoff data? I'd love to see the xGF results for the games

 

[Zil]

Hayes' game has really dropped. And that Staal-Holden pairing is just blech.

 

[TheTakedown]

Jesus... a good game by girardi and he's still hemorrhaging shot attempts like crazy

 

[eperry]

Why doesn't Manny scrap playoff data? I'd love to see the xGF results for the games

I had to touch up some code. It should be working now.

 

[Ola]

Hayes' game has really dropped. And that Staal-Holden pairing is just blech.

They aren't giving up many goals though and they do create a lot top notch chances, although they have been a bit snake bitten lately. Think especially Miller and Hayes can pick it up a bit though, guys like Grabner are always by nature streaky. But Hayes and Miller haven't scored much by will either. They have been caught a bit too much in the the looking for Hail Marys to Grabner game. They probably have to step up and play really good hockey for this team to go anywhere. We are pretty caught in the strategy we have now wth Hayes taking a lot of D zone FOs for Grabner to get ice to skate on when they head up ice. Maybe AV should try adjusting a bit.

 

[Ola]

Jesus... a good game by girardi and he's still hemorrhaging shot attempts like crazy

Does anyone provide info on SAs per period?

Some of these are coming late in the game, and they did face Rads and MaxPac as much as possible, but it also seems like they had some problems against the Habs second (third?) line with some ugly shifts. With Girardi though, it is important to take into account that when they get deep, he very much still play Torts hockey. Giving up a lot of SAs but still keeping things mostly to the outside, getting into lanes. Check the Sv% in relation to SAs by CBJs Ds, most of their blueline is in the top 30 in the league.

It's when they give up a lot of SAs AND starts losing players the actual results goes South too... For Girardi to warrant a spot in the NHL though, it still just comes down to him being a great PKer and being able to take all those toughest late game shifts and still be able to walk off the ice a winner. For a long time he did that, we had the best records in the league going into the 3rd with a lead for long stretches. Is he still capable of that? We will know soon enough.

 

[Ola]

AV is really sheltering Staal and Holden, from last game:

Name/CF%
Marc*Staal 77.42
Nick*Holden 74.07
Mats*Zuccarello 69.23
J.T.*Miller 66.67
Chris*Kreider 63.64
Oscar*Lindberg 63.64
Michael*Grabner 63.16
Mika*Zibanejad 63.16
Kevin*Hayes 59.09
Brady*Skjei 55.88
Pavel*Buchnevich 54.55
Brendan*Smith 54.05
Rick*Nash 51.85
Derek*Stepan 50.00
Jesper*Fast 50.00
Jimmy*Vesey 48.28
Dan*Girardi 44.83
Ryan*McDonagh 41.38

http://naturalstattrick.com/game.php?season=20162017&game=30114

 

[Inferno]

Marc Staal played more minutes than Skjei or McDonagh....what....the....frak.

 

[haohmaru]

Marc Staal played more minutes than Skjei or McDonagh....what....the....frak.

Inaccurate. McD played nearly 3 minutes longer than Staal.

 

[Inferno]

Inaccurate. McD played nearly 3 minutes longer than Staal.

i must be looking at the wrong TOI then? i was looking at naturalstattricks site...does that do only EV or something?

 

[Mac n Gs]

i must be looking at the wrong TOI then? i was looking at naturalstattricks site...does that do only EV or something?

Unless you change the filter, yeah it represents 5v5 TOI first. Look at you using fancy stats

 

[Irishguy42]

i must be looking at the wrong TOI then? i was looking at naturalstattricks site...does that do only EV or something?

NaturalStatTrick defaults to 5v5 for everything, but it can be changed.

 

[Inferno]

ok gotcha, NM then...nothing to see here!

 

[silverfish]

Long read, but a good read: http://www.newyorker.com/magazine/2017/02/27/why-facts-dont-change-our-minds

Even after the evidence “for their beliefs has been totally refuted, people fail to make appropriate revisions in those beliefs,†the researchers noted. In this case, the failure was “particularly impressive,†since two data points would never have been enough information to generalize from.

Surveys on many other issues have yielded similarly dismaying results. “As a rule, strong feelings about issues do not emerge from deep understanding,†Sloman and Fernbach write. And here our dependence on other minds reinforces the problem. If your position on, say, the Affordable Care Act is baseless and I rely on it, then my opinion is also baseless. When I talk to Tom and he decides he agrees with me, his opinion is also baseless, but now that the three of us concur we feel that much more smug about our views.