The Advanced Stats Thread Episode IX

[Filthy Dangles]

Another issue is the fluidity of the game, and the ambiguity of possession and if certain events are classified as 'offense' or 'defense'. You could say any situation with the puck is offense and without it is defense but that's over-simplistic IMO, for the reasons above. Let me give you some examples....

Is retrieving the puck down low and evading a modeate-heavy forecheck offense or defense? Some combo of both...?

Pouncing on a loose puck in the offensive zone or stepping up at the offensive blue line and keeping a cycle or possession alive? What is that?

It happened in the offensive zone when your team might have had the puck for most of that sequence but then they didn't. Did it become 'defense' the instant the team lost the puck? Can you "play defense in the offensive zone"?

 

[silverfish]

Another issue is the fluidity of the game, and the ambiguity of possession and if certain events are classified as 'offense' or 'defense'. You could say any situation with the puck is offense and without it is defense but that's over-simplistic IMO, for the reasons above. Let me give you some examples....

Is retrieving the puck down low and evading a modeate-heavy forecheck offense or defense? Some combo of both...?

Pouncing on a loose puck in the offensive zone or stepping up at the offensive blue line and keeping a cycle or possession alive? What is that?

It happened in the offensive zone when your team might have had the puck for most of that sequence but then they didn't. Did it become 'defense' the instant the team lost the puck? Can you "play defense in the offensive zone"?

It's subjective. I'd say:

In defensive zone without puck: defense
In defensive zone with puck: transition offense
In neutral zone with puck: transition offense
In neutral zone without puck: transition defense
In offensive zone without puck: transition defense
In offensive zone with puck: offense

 

[silverfish]

 

[1Knee1T]

 

[GeorgeKaplan]

CC: @silverfish

 

[nyr__1994]

CC: @silverfish

There is a lot of great information in here. Long read, but very enlightening. I am someone that hasn't gone all in on the shot volume analytics. One of the main reasons for that is that not all shots are created equal, even if they are coming from the same spot on the ice. What Vally is doing is creating buckets within each shot location to provide more insight. A very manual process, but the information is starting to get there.

After reading this, AV's Rangers at their peak, while getting out shot attempted, I would guess that the NYR expected scoring percentage was higher due to the nature of the shots they did take. While they weren't as numerous as the opposition, they were from areas of the ice and on plays that had a higher chance of going in.

Hopefully that makes sense. But what seemed to happen in the playoffs is that teams tighten up and those higher danger chances are not available as often.

 

[Mac n Gs]

There is a lot of great information in here. Long read, but very enlightening. I am someone that hasn't gone all in on the shot volume analytics. One of the main reasons for that is that not all shots are created equal, even if they are coming from the same spot on the ice. What Vally is doing is creating buckets within each shot location to provide more insight. A very manual process, but the information is starting to get there.

After reading this, AV's Rangers at their peak, while getting out shot attempted, I would guess that the NYR expected scoring percentage was higher due to the nature of the shots they did take. While they weren't as numerous as the opposition, they were from areas of the ice and on plays that had a higher chance of going in.

Hopefully that makes sense. But what seemed to happen in the playoffs is that teams tighten up and those higher danger chances are not available as often.

AV’s teams were great offensively. The offense never was the problem, it was always the defense yielding high danger chances because teams knew how to exploit our weakness

 

[SnowblindNYR]

So, we can't attach excel spreadsheets on this site. But I've been working on a Fantasy football forecast model for about a week now and it's pretty good. It's not more accurate than just taking the average but it's within 1 percent of accuracy (in total for all QBs). I suspect any model would be close to just taking an average since most models are built on averages from one standpoint or another.

Category	Fill-In	Notes
QB Full Name (Pre-2017)	Cam Newton
Home/Away	Home
Dome/Open	Dome
Inclement/Clear	Clear	Inclement is Precipitation or Fog; For Dome put Clear
Def Pass Yards/Gm	200
Def Rush Yards/Gm	100
Temp (F)	71	For Dome put 71
Wind (MPH)	0	For Dome put 0
Simple Average Estimate	21.92
Model Estimate	18.93

[TBODY] [/TBODY]

The first number is just taking an average, the second is based off my model.

My methodology was just multiple regression for the variables listed above and then I came up with an index of how each QB does on average compared to what he's expected to do based on the conditions and the model. The model result is index*regression equation.

 

[SnowblindNYR]

It's really football, but I guess it's an analytics thread and I figured I'd share it with my fellow nerds. Here's my sports analytics blog. I did hockey analysis back in the day but am too lazy to go back and write a blog based on analysis from like 2 years ago. I actually shared it back then.

Analysis

 

[silverfish]

CC: @silverfish

Do not subscribe tot his. Who wants to give me the 'saved you a click?'

I think I'm two games behind on chart dumps, here?

 

[Machinehead]

I don't think tonight's game was as bad as it's being made out to be. Buffalo had like 74 powerplays.

 

[Blue Blooded]

It's interesting that we seem to perceive the team as having a good PP and an awful PK, but the metrics say otherwise. The Rangers are currently 26th in 5v4 xGF/60 and 16th in 4v5 xGA/60. Staal-Pionk have been legitimately bad 4v5 but Skjei and Smith have been great so far and made up for it. Meanwhile the team is below average at generating shot attempts 5v4 and the quality of those attempts are very poor (26th in 5v4 ixFSh%)

The PDO effects have also been small and the results in terms of goals are average (5v4) to slightly below average (4v5).

 

[Machinehead]

It's interesting that we seem to perceive the team as having a good PP and an awful PK, but the metrics say otherwise. The Rangers are currently 26th in 5v4 xGF/60 and 16th in 4v5 xGA/60. Staal-Pionk have been legitimately bad 4v5 but Skjei and Smith have been great so far and made up for it. Meanwhile the team is below average at generating shot attempts 5v4 and the quality of those attempts are very poor (26th in 5v4 ixFSh%)

The PDO effects have also been small and the results in terms of goals are average (5v4) to slightly below average (4v5).

Maybe our PP stinks because we're playing Neal Pionk at the point over the best PP pointman of the last 10 years.

/KermitSippingTea.mp3

 

[silverfish]

Maybe our PP stinks because we're playing Neal Pionk at the point over the best PP pointman of the last 10 years.

/KermitSippingTea.mp3

It's an imperfect science based on small sample sizes, but using Zibanejad as a proxy for PP1...

W/ Shattenkirk: 24:07, 114.44 shot attempts per 60, 17.42 high-danger chances per 60
W/ Pionk: 17:55, 83.72 shot attempts per 60, 16.74 high-danger chances per 60

http://naturalstattrick.com/playerr...2&sit=pp&stdoi=oi&rate=y&v=t&playerid=8476459

I mean, it is not a coincidence that from 15-16 through 17-18, among d-men who logged at least 100 minutes on the PP, Shattenkirk is second in points per 60.

Player Season Totals - Natural Stat Trick

 

[Blue Blooded]

It's an imperfect science based on small sample sizes, but using Zibanejad as a proxy for PP1...

W/ Shattenkirk: 24:07, 114.44 shot attempts per 60, 17.42 high-danger chances per 60
W/ Pionk: 17:55, 83.72 shot attempts per 60, 16.74 high-danger chances per 60

http://naturalstattrick.com/playerr...2&sit=pp&stdoi=oi&rate=y&v=t&playerid=8476459

I mean, it is not a coincidence that from 15-16 through 17-18, among d-men who logged at least 100 minutes on the PP, Shattenkirk is second in points per 60.

Player Season Totals - Natural Stat Trick

The 5-year sample (13-18) has Shattenkirk as 6th overall, including forwards, among players with 250+ PP minutes (chosen for 50/season) and 1st among defencemen by a mile (the next defenceman John Carlson is 25th). The five players above Shattenkirk are: Bäckström, Marner, Malkin, Giroux, and Kucherov. And the difference to the players below is pretty steep; Shatty scored 6.47 P/60 while e.g. Tavares scored 5.76 and Ovechkin 5.45.

Oh, and he did so with little help. The next Blue on the list is Alex Steen in 52nd. The guys ahead of him all had teammates in the top-20.

Top-5 PPQB is probably an unfairly conservative estimate of Shattenkirk.

 

[lilphildub]

Yea I think its obvious Pionk has ran his course. Time to give Shatty PP1 minutes and try DeAngelo on the second unit

 

[Harbour Dog]

I generally don't check Corsica during a game, but I had a look at about the 16 minutes mark of that period out of curiosity.

Our all-situations xGF was 0.76 at the 16-minute mark. That's ludicrous. Kreider's goal alone was a >76% chance of scoring in my opinion, and that's not to mention McLeod and Vesey's chances.

I'm going to pay more attention to it during games from now on, but that number alone is enough for me to think that I've been giving too much credit to xG stats.

 

[silverfish]

I generally don't check Corsica during a game, but I had a look at about the 16 minutes mark of that period out of curiosity.

Our all-situations xGF was 0.76 at the 16-minute mark. That's ludicrous. Kreider's goal alone was a >76% chance of scoring in my opinion, and that's not to mention McLeod and Vesey's chances.

I'm going to pay more attention to it during games from now on, but that number alone is enough for me to think that I've been giving too much credit to xG stats.

The outputs are only as good as the inputs. NHL is listing the Kreider goal as a 10 foot wrist shot that isn't a rush attempt. If you take a magnifying glass to xG, you'll be able to find countless items like this that are "wrong" (for lack of a better term). It's use is when viewing it as a whole, not as an individual.

 

[Leetch3]

can one of you fine gentleman show me a comparison of vesey's advanced stats this year compared to last year? eye test he's alot better this year and curious if the stats tell the same story.

 

[Harbour Dog]

The outputs are only as good as the inputs. NHL is listing the Kreider goal as a 10 foot wrist shot that isn't a rush attempt. If you take a magnifying glass to xG, you'll be able to find countless items like this that are "wrong" (for lack of a better term). It's use is when viewing it as a whole, not as an individual.

Sorry, would of responded sooner but I never got a notification for some reason.

I get this, and I assume that the hope is it balances out in the long run; but it highlights a bias that I hadn't previously considered that xG had.

Players that take more risks to create tap-in goals that are practical empty netters, only get credited with being on the ice for a "wrist shot" most of the time.

Just did a quick check of Hayes and Zucc (being two guys that may be losing out to this concept): since 2015, they've outperformed their xGF% by 5.22% and 2.5% respectively, with PDO accounting for only 2.3% and 1.2% of each. And that isn't surprising.

I'm legit probably going to start adding 2-3% to their xGF%s when I read them from now on.

 

[silverfish]

Sorry, would of responded sooner but I never got a notification for some reason.

I get this, and I assume that the hope is it balances out in the long run; but it highlights a bias that I hadn't previously considered that xG had.

Players that take more risks to create tap-in goals that are practical empty netters, only get credited with being on the ice for a "wrist shot" most of the time.

Just did a quick check of Hayes and Zucc (being two guys that may be losing out to this concept): since 2015, they've outperformed their xGF% by 5.22% and 2.5% respectively, with PDO accounting for only 2.3% and 1.2% of each. And that isn't surprising.

I'm legit probably going to start adding 2-3% to their xGF%s when I read them from now on.

But if you're adding to players, you have to take away from other players, and thus, you're back to where you started

 

[Irishguy42]

can one of you fine gentleman show me a comparison of vesey's advanced stats this year compared to last year? eye test he's alot better this year and curious if the stats tell the same story.

I will get back to this when I'm at a laptop or desktop. Or if someone else gets to it first

 

[Harbour Dog]

But if you're adding to players, you have to take away from other players, and thus, you're back to where you started

His xG model balances out exactly to actual G?

I'm most concerned with using xGF% for individual player assessment anyway, so I'm definitely going to do some adding and subtracting. It's gonna be a little more labourious, but I won't be able to help myself.

 

[silverfish]

His xG model balances out exactly to actual G?

No, and it shouldn't, because goalies. xG is goalie independent and G is goalie dependent. But if you're saying that you think you need to add 3% to Kreider because his chances were higher than you thought, you need to deduct 3% from whoever was on the ice against him.

 

[Harbour Dog]

No, and it shouldn't, because goalies. xG is goalie independent and G is goalie dependent. But if you're saying that you think you need to add 3% to Kreider because his chances were higher than you thought, you need to deduct 3% from whoever was on the ice against him.

Oh yeah, I completely agree.

On a game-by-game basis, any additions or subtractions I would make would be entirely subjective and worth very little anyway.

I'm thinking in terms of larger sample sizes. Like, if Hayes' Corsica xGF% is 50 by Xmas, and his PDO is in line with his average since 2015 (102.3), then perhaps it would be more accurate to estimate that his GF% will be 55.22% as opposed to the 51.15% (?) that those numbers suggest.