2017-18 stats and underlying metrics thread [Mod: updated season]

Status
Not open for further replies.

Whileee

Registered User
May 29, 2010
46,061
33,043
I love how people want to destroy +/- as having any usefulness at all. That is simply hogwash but the fancy stats movement didn't invent it so it sucks.

Plus minus is flawed. Corsi is flawed, fenwick is flawed WAR is flawed, eye test is flawed.

If hockey was an easy sport to measure, someone would have invented effective measurement of it.

Hockey is a wonderful beast, magical and hard to measure. The key I think is to use all the different methods possible to get an overall view and FORM your own opinon.

I use myself as a measure in my 'hockey career' :laugh:
I have always been a defensive defenseman. I pride myself in shutting down the other team's best players.
I feel like I am decent at transitioning the puck out of my zone and very good at recovering it.
I always measure my plus minus, because it matters to me. I want to be on the ice for more goals for than against (honestly, I just don't want to be on the ice for goals against). I believe in preventing first, and then creating.

If a player demonstrated a plus rating year over year, on different teams, in different scenarios or with different partners, it's safe to assume that he is driving wins for that team. I know plus minus depends on circumstance and who you play with, and SOME people either benefit or suffer from that, but at the end of the day the stat still has value and THAT is why the NHL still uses it.

JMHO

There's an easy improvement to the +/- stat. It's the 5v5 GF%. This is essentially a plus/minus but eliminates special teams and extra man situations (like pluses for short-handed goals or empty-net goals). It's still a bit of an unstable stat because there is still "luck" involved, though I think it has too often been overstated because I don't think a lot of "advanced" stats accurately track the danger level of shots again and the factors that lead to goals against. As an example, I think the Jets have been much better at preventing those top-end chances that led to so many goals last season. So, some of the reduced goals against is better goaltending, but I also think that they are better at systematically protecting the danger areas in front of the net and bang-bang plays. They are also pretty good at generating good shots from the slot area. This is illustrated in the following two graphics.

WPG


WPG


Now look at the Jets shot generation by location in the previous season. It's clear that they are better this season in generating shots from the slot area than they were last season.

WPG
 

JetsFan815

Registered User
Jan 16, 2012
19,122
23,879
I love how people want to destroy +/- as having any usefulness at all. That is simply hogwash but the fancy stats movement didn't invent it so it sucks.

Plus minus is flawed. Corsi is flawed, fenwick is flawed WAR is flawed, eye test is flawed.

If hockey was an easy sport to measure, someone would have invented effective measurement of it.

Hockey is a wonderful beast, magical and hard to measure. The key I think is to use all the different methods possible to get an overall view and FORM your own opinon.

Witch doctors are bad, don't have the ability to diagnose and treat basic medical conditions and lead to deaths of many in their communities. Homeopathic doctors are pretty much quacks and if used to treat people with serious illnesses will lead to deaths. University educated classical doctors sometimes have bad diagnosis and lead to deaths but otherwise do a fantastic job of saving the lives of millions every year.

All 3 kinds of doctors have their flaws and therefore should be put in the same class and treated with the same amount of respect :sarcasm:
 

Saidin

Registered User
Mar 18, 2015
1,250
1,043
WHOA. Potential sea change alert re WPG's penalty kill:
Full season CA/60: 31st
Full season HDCA/60: 17th
Since Dec 1 CA/60: 15th
Since Dec 1 HDCA/60: 3rd

I'm not sure how to link a tweet in here, but I got this from a guy Westwood retweeted 'Murat'
 

Joe Hallenback

Moderator
Mar 4, 2005
15,358
21,345
I would be curious to know what fancy stats say about Anton Stralman. He leads the league in +/- and he isn't a prolific scorer by any means. In reverse Ekman-Larsson leads the league in - by a lot.
 

surixon

Registered User
Jul 12, 2003
48,728
69,078
Winnipeg
I would be curious to know what fancy stats say about Anton Stralman. He leads the league in +/- and he isn't a prolific scorer by any means. In reverse Ekman-Larsson leads the league in - by a lot.

Stralman is seen in a pretty positive light by most metrics if memory serves me correct. He's a very good and underrated player.
 

Whileee

Registered User
May 29, 2010
46,061
33,043
I would be curious to know what fancy stats say about Anton Stralman. He leads the league in +/- and he isn't a prolific scorer by any means. In reverse Ekman-Larsson leads the league in - by a lot.
2015/16 through 2017/18 (to date)...

From Corsica (5v5, venue and score adjusted):

Points/60 0.71
RelCF% +1.1
RelxGF% +2.35

Low-scoring, but effective defenseman. I think he would fit in very well alongside Morrissey.

Main issue is that he only has this season and next before UFA, so he'd be a short-term solution for the Jets unless they could extend him. He'll be 32 starting next season, so a bit of a risk to think about extending him. I think the Jets would need to receive very good futures in addition to Stralman, but he could fill in for Trouba for next season.
 

Halberdier

Registered User
May 14, 2016
4,467
4,980
There's an easy improvement to the +/- stat. It's the 5v5 GF%. This is essentially a plus/minus but eliminates special teams and extra man situations (like pluses for short-handed goals or empty-net goals). It's still a bit of an unstable stat because there is still "luck" involved, though I think it has too often been overstated because I don't think a lot of "advanced" stats accurately track the danger level of shots again and the factors that lead to goals against. As an example, I think the Jets have been much better at preventing those top-end chances that led to so many goals last season. So, some of the reduced goals against is better goaltending, but I also think that they are better at systematically protecting the danger areas in front of the net and bang-bang plays. They are also pretty good at generating good shots from the slot area. This is illustrated in the following two graphics.

WPG


WPG


Now look at the Jets shot generation by location in the previous season. It's clear that they are better this season in generating shots from the slot area than they were last season.

WPG

Excellent post you have there.

I'm not opposing using just 5-on-5 Gdiff, as that is at least as fair comparison as it gets regarding 5-on-5 play, doh. However, I "developed" just for fun a #+/- stat. C# programmers maybe gets the joke (C -> C++ -> C++++ == C#).

I think you really should take into account goals against when on PP and goals for when on PK. But also as every goal count in the end, to some extend you should take into account all the goals, all situations, with some weighting. There could be all kinds of weighting, but I decided to use 0.5 or 50% for PP GF and also the same for PK GA to keep them symmetrical. Then I added 2.0 or 200% weight for PP GA and the same for PK GF to equalize things. Even strength GF and GA counts as 1 as always. So while all situations raw GF-GA would have weighting as 1.0 (ES), 1.0 (PP GA, PK GF) and 1.0 (PK GA, PP GF), normal +/- has weighting as 1.0, 1.0 and 0.0 respectively, my #+/- model has weighting of 1.0, 2.0 and 0.5 respectively.

As can be seen from the results, in most of the cases this system is a bit too easy on PP players and a bit too harsh on PK players, but there are cases when active PK players can get even, as is the case with Myers who happens to have exactly 3 PKGF (+3*2.0 = +6.0) and 12 PKGA (-12*0.5 = -6.0), having PK#+/- 0.0.

I'd say this is still more fair system than the current +/- which just completely ignores PKGA and PPGF, but still adds in PKGF and PPGA.

Please find an attached pdf file with "#+/-" statistics for the Jets this season. I'm not all that surprised about those guys that did came on top of the list, and the bottom end is not a surprise as well. Also that all situations GF-GA is visible, and while totally unfair, no surprises here either.
 

Attachments

  • WPG#+-.pdf
    31.2 KB · Views: 10

Aavco Cup

"I can make you cry in this room"
Sep 5, 2013
37,630
10,440
Dellow also mentions in his piece how the Jets have turned into a middle of the pack PK team in terms of CA/60
 

CorgisPer60

Barking at the net
Apr 15, 2012
21,203
9,394
Please Understand
WHOA. Potential sea change alert re WPG's penalty kill:
Full season CA/60: 31st
Full season HDCA/60: 17th
Since Dec 1 CA/60: 15th
Since Dec 1 HDCA/60: 3rd

I'm not sure how to link a tweet in here, but I got this from a guy Westwood retweeted 'Murat'

'Murat' is the local scribe for The Athletic, so anything he tweets out will be pretty legitimate.
 

Saidin

Registered User
Mar 18, 2015
1,250
1,043
'Murat' is the local scribe for The Athletic, so anything he tweets out will be pretty legitimate.

Ok well, then that's good. So much doom and gloom early in the season about or PK, but from what he's saying it's drastically has improved.
 

garret9

AKA#VitoCorrelationi
Mar 31, 2012
21,738
4,380
Vancouver
www.hockey-graphs.com
Corsi is flawed
Eye test is flawed
WAR is flawed
etc.

But two things...

ONE

+/- is not flawed but useless
It doesn't even show outscoring properly...
The only thing +/- shows is SPECIFICALLY +/-
In 2014-2015 Mark Stuart had the worst EV goal differential, PK goal differential, and PP goal differential for the team's defenders. BUT, he was middle of the pack for +/- because of the arbitrary removal of some goals.

TWO

Not all things are equally flawed.
The reason why everything is flawed is because each thing only shows a part of the picture. That said, each thing does not show the same amount of the full picture. There is also overlapping portions of the picture, and there can be outside influences as well.
That's why combination is best method, but there is an importance to weight things appropriately.
 

raideralex99

Whiteout Is Coming.
Dec 18, 2015
4,750
9,014
West Coast
I'm not trying to be a jerk but Corsi is useless ... too many variables. You have 12 players on the ice 2 refs and a schedule maker and you are trying to tell me which player has better stats? Teams play different with a lead ... teams play different with certain refs ... teams play different playing back to back games. Those are the stats that you should be more concern with. Keeping player stats is only useful in baseball because it's one on one why do you think the MLB all-star game is the only all-star game that is exciting to watch the other all-star games are boring and just a cash cow for the leagues.
 

Halberdier

Registered User
May 14, 2016
4,467
4,980
@raideralex99

Usefulness of Corsi is at least seriously limited, though it might be good tool for the use it was originally intended to. Measuring the workload of goalies as they need to react to every shot and not only those that they did catch.

Using Corsi as a measure for player A being better than player B is really questionable, as the causality isn't there and the statistical average correlation between Corsi and real goals is minimal. There have been bold (though false) claims for Corsi such as that "Corsi + luck can explain 90% of the success", that was also in one of the linked articles in OP.

Let's say Whilee got it right with his analysis, and I did understand his analysis correctly, there is only 0.18 correlation between raw Corsi and real goals. Combined with above false claim by "90% can be explained by Corsi and luck" that leaves 72% factor for pure luck, and sorry, but I don't buy it at all. Guys that are great in winning are not great because they are lucky. In fact, in larger sample sizes the factor for lucky pretty much disappears, unless we are talking about career altering injuries and such.

Guys that are are playing identical roles with identical teams and identical lineups, they are 100% comparable by looking just one stat such as +/-, and just tell that this guy is better than the another. To be able to do that with Corsi, you need to also take into account that they have to be with identical talent and identical play. For direct comparison between players Corsi is therefore even worse than faulty +/-.
 

Joe Hallenback

Moderator
Mar 4, 2005
15,358
21,345
I hate when people use Corsi as a possession stat. Too me that is just wrong. Having the puck is posession, shooting it towards the net is something entirely else. I think Corsi has value but not as a possession stat.
 
  • Like
Reactions: Halberdier

Halberdier

Registered User
May 14, 2016
4,467
4,980
I hate when people use Corsi as a possession stat. Too me that is just wrong. Having the puck is posession, shooting it towards the net is something entirely else. I think Corsi has value but not as a possession stat.

Exactly. Only thing Corsi tells about the possession is that at least amount of Corsi your team has lost the possession (at least briefly).
 

Halberdier

Registered User
May 14, 2016
4,467
4,980
Corsi measures useful possession

It doesn't.

Corsi doesn't even measure useful shots, let alone possession.

There are useful measurements for measuring possession, and all of them measure time with puck or time in the o-zone.

Counting successful passes could be also a way to measure possession. I know that such statistics are being done by some.
 

csk

Registered User
Nov 5, 2015
2,682
269
Winnipeg, MB
It doesn't.

Corsi doesn't even measure useful shots, let alone possession.

There are useful measurements for measuring possession, and all of them measure time with puck or time in the o-zone.

Counting successful passes could be also a way to measure possession. I know that such statistics are being done by some.

All shots are useful, just to varying degrees. This is why stats like xG exist, to try to measure how useful each shot is.
 

Halberdier

Registered User
May 14, 2016
4,467
4,980
All shots are useful, just to varying degrees. This is why stats like xG exist, to try to measure how useful each shot is.

Their usefulness varies from 0.00 to 1.00, so I think it's bit of a stretch to say all shots are useful.
 

Halberdier

Registered User
May 14, 2016
4,467
4,980
There is a nonzero chance of a shot going in from any location.

Nonzero might be 0.00, easily. There is also nonzero change for any puck that is passed on opposite side of the rink or a puck that is touched by skate or a referee to go in. These are not Corsi events at all.

There is also nonzero probability for all propositional claims (most of the arguments you can imagine in HFBoards) to be truthful, but that doesn't make them truthful.

There is a nonzero change than the Moon is made of cheese. I would still not say it's truth that the Moon is made of cheese.
 

Joe Hallenback

Moderator
Mar 4, 2005
15,358
21,345
Corsi measures useful possession

Players come down on a 3 on 2, Player with the puck down the wing shoots the puck, misses the net and bounces back out of the zone and the opposing team takes it and comes down 3 on 2 and scores. How is shooting the puck in the situation "useful". I would think that missing the net there was detrimental to the team and not useful and you gave up possession.

I do however get the correlation that shooting the puck move often does result in a better chance of scoring. That is just one of the oldest things in the books though "You don't score on every shot you don't take".
 
Status
Not open for further replies.

Ad

Upcoming events

Ad

Ad

-->