Ongoing Stats and Analytical Discussion Thread: Battle of the Defense

Feed Me A Stray Cat

Registered User
Mar 27, 2005
14,847
144
Boston, MA
I think it's reasonable to assume that the Devils' forward corps is skilled at converting chances, and thus they can score on a lower amount of shots. However they're probably not going to sustain 9.7% ES shooting.

However, unless they clean up their defensive play, they're going to get hammered soon.
 
  • Like
Reactions: Bleedred

Devils731

Registered User
Jun 23, 2008
12,199
16,108


There's an interesting thread there where he analyzes the Devils up to now. In short, Devils are 29th/30th in unblocked and regular shot rates adjusted for score, allowing more shots in high danger areas (shown below), rank 5th in 5-on-5 save percentage, and have the ~26th worst penalty differential.



With the shot heat chart. The shots very close in come mostly from the left side. I wonder if we would see shots from that close in are not so unfavorable if it’s on the goalies glove side.

I guess you could take Cory’s historical goals allowed chart and compare it to the Devils heat shot allowed chart and see if there seems to be a correlation that either picture alone wouldn’t suggest.
 

Wingman77

Registered User
Mar 16, 2010
20,251
765
its the total opposite in this case. Go look at some of BG's posts from late last night.

He decided that ALL situations should be used, not only 5 on 5....because he said so. He also demeaned people and said they don't understand statistics like he does.

Oh I'm not referencing the particular disagreement that has been ongoing. Just more so in general. I can see why these stats can be off-turning to people. It's not even the stats that turn people off, it's some of the people incessantly behind the stats that turn people off.

Personally I think there is the need for such stats in moderation. It gives fans another way to look at situations and more importantly, those involved with the daily decisions and that are actually running teams, other options to look into why something may or may not be working.
 

devilsblood

Registered User
Mar 10, 2010
29,442
11,701
its the total opposite in this case. Go look at some of BG's posts from late last night.

He decided that ALL situations should be used, not only 5 on 5....because he said so. He also demeaned people and said they don't understand statistics like he does.
To be fair, he does raise an interesting point. Why not all situations corsi? Why is it less helpful then 5v5?
 

Zippy316

aka Zippo
Aug 17, 2012
19,485
4,473
New Jersey
To be fair, he does raise an interesting point. Why not all situations corsi? Why is it less helpful then 5v5?

Because PP and PK is a completely different game. The sample size is so much smaller over the course of a full season and PP units change so much that it's hard to quantify individual or team effects on Corsi.

The game is primarily played 5-on-5 which is why that's used. If you win consistently win the battle there, you should win the game unless you have an unusually high SH% or unusually low SV%.
 

devilsblood

Registered User
Mar 10, 2010
29,442
11,701
Because PP and PK is a completely different game. The sample size is so much smaller over the course of a full season and PP units change so much that it's hard to quantify individual or team effects on Corsi.

The game is primarily played 5-on-5 which is why that's used. If you win consistently win the battle there, you should win the game unless you have an unusually high SH% or unusually low SV%.
I don't know if that answers the question though.

Completely different game? OK but why are they lesser indicators.

Sample size is smaller? But they are big enough to change the corsi stat.

Units change? Not sure why that's relevant when different lineups don't have that effect on 5v5 corsi. It's about totals, not the details of who is shooting them.
 

The Devil In I

Registered User
Jun 28, 2005
4,179
1,121
Chicago
I don't know if that answers the question though.

Completely different game? OK but why are they lesser indicators.

Sample size is smaller? But they are big enough to change the corsi stat.

Units change? Not sure why that's relevant when different lineups don't have that effect on 5v5 corsi. It's about totals, not the details of who is shooting them.

Teams don't play an equal proportion of time on the PP and PK. It's 100% even across the board when only using 5 on 5 play. Also, I wouldn't be surprised if the change in team corsi from including it just resulted in showing which team generally got a higher ratio of PP/PK time. Which isn't even necessarily a good indicator of which teams drew more penalties or had better PP/PK units. PP/PK time is just too different from standard 5 on 5 play to be a helpful addition.
 

Feed Me A Stray Cat

Registered User
Mar 27, 2005
14,847
144
Boston, MA
I have seen some arguments from smart people that say we should look at an all-situations Corsi, since it gives a larger sample of data and rewards teams that take less penalties and draw more. But it is still useful to isolate the performance of 5v5, PP, PK, etc., to see how each team fares in each area.

Another thing to think about: only a small percentage of a team's players play on the PP. Same with PK. Looking at 5v5 Corsi gives a better indicator of overall team quality.
 
  • Like
Reactions: Bleedred

guitarguyvic

Registered User
Mar 31, 2010
8,680
6,797
We need advanced analytics to know that the Devils give up a ridiculous amount of high quality chances? It's pretty obvious just from watching the games, the shots against totals, the number of times we've been outshot, etc.If it doesn't improve this is easily a 7-10 seed team by the end of the season.

Toronto got quality chances during last night's game, but it was still a much better team effort on D and they actually outshot the opposition. It gives me some hope that it get cleaned up to some degree, which would go a long way towards this being a playoff team.
 

None Shall Pass

Dano moisturizes
Jul 7, 2007
15,379
11,597
Brooklyn
I don't know if that answers the question though.

Completely different game? OK but why are they lesser indicators.

Sample size is smaller? But they are big enough to change the corsi stat.

Units change? Not sure why that's relevant when different lineups don't have that effect on 5v5 corsi. It's about totals, not the details of who is shooting them.

The question is based on the premise that Corsi is an end-all-be-all stat, which it isn't and shouldn't be used as.

Penalties are often random events, and the amount of events can change by the game. You can go a game having one PP, then have eight the next game. Meanwhile, even if you only have one PP, you can also go on the PK eight times in that same game. And if you're on the PK, your total Corsi takes a hit, because you have less personnel to actually control possession.

And let's look at that a little closer. Let's say your team has 1 PP and 8 PK in one game. Assuming there's no scoring on those plays, that's 16 total minutes of lopsided possession against you and 2 minutes for you. 18 total minutes of non-5-on-5 hockey, which means there's 18 minutes of statistically weighted possession favorability. Based on random chance/events (The referee's discretion).

And then there's 42 minutes of even strength (Usually 5-on-5). This is a much larger sample, and it is a natural, unaltered sample. It's evenly weighted for the two teams - Each has five skaters and a goalie.

If each team has the same amount of PP-time and PK-time in a game, it might be more useful to do all-situation Corsi. But in lopsided games, it paints an unfair picture on possession.

And Corsi doesn't predict it who will win a game - it just shows who possessed the puck more. And teams who possess the puck more tend to have more scoring chances and more scoring, which tends to lead to wins. But it's not always the case, as with almost every statistic.
 

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
To be fair, he does raise an interesting point. Why not all situations corsi? Why is it less helpful then 5v5?

It's not really "interesting", it's metaphysically logical.

Either the metric works as a predictive value for team success, or it does not work as a predictive value for team success.

To exclude special teams from the broader data set is completely illogical.

The implied logic behind a Team's Corsi rating doesn't go away simply because you're on the PP or PK, it still applies. To argue otherwise, ironically means you do not believe in the core logic behind Corsi.

The fact that PP and PK time are "much smaller" data sets as people are noting is contextually completely irrelevant. Taken together, they BOLSTER your net available data, even if combined they may only represent something like 15% of the total data.
 
Last edited:

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
I once found a statistician's write-up on Corsi (I just Googled, couldn't find it again to post it here), and while he was open-minded to Corsi he had his doubts, so he crunched some data in the proper fashion, complete with p-values.

Before doing so, however, he tried to look at the genesis and early days of Corsi to get a better understanding of the original intent of the metric, and like me, he could not logically understand why "5v5 Corsi" became a thing with respect to Team Corsi, because like me he also noted that doesn't really make sense. Net/net, he basically couldn't figure out how it evolved that way and basically decided it was just because, (paraphrasing) "conventionally that's just what most people seemed to initially do".

Anyway, his conclusion after his analysis, was that he determined that as a predictive metric, this "advanced statistic" was no better (and actually worse) than.....uhhhh.... "really simplistic statistics" for lack of a better term, such as just looking at goal differential.

In the 3 seasons he conducted the study, uber simplistic GF - GA was better predictive of NHL team's future success than 5v5 Corsi.

And I wasn't surprised in the least bit.
 

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
Oh, almost forgot to mention in my 5v5 "essay" above, it's also completely illogical that with 5v5 Corsi you're deleting all data for 4v4 play.

You should especially believe that that's bizarre if you're going to be one of the, "only even strength play" matters people.

Ummm....... okay, I'll bite, then why didn't the gifted stat geniuses make it Even Strength Corsi rather than 5v5 Corsi?

IIRC, a few minutes on average of an NHL game are 4v4, which is roughly 3.3% of all ES data "lost" for no other logical reason than, well...... we've just always calculated it this other way, so............ uhhh....we kind of forgot about that.
 

Devils731

Registered User
Jun 23, 2008
12,199
16,108
BG, you keep saying using only even strength is illogical but you’ve been given reasons why people exclude it. You may disagree but excluding PP and PK isn’t being done without thought.

Imagine if I kept ranting it’s illogical to not include all special team goals in a players +/- because you’re not accurately reflecting the goals that happen while a player is on the ice. People explain to me why they’re excluded but I just ignore their explanations and reiterate why I think it’s illogical. Can you see the illogic of how a discussion like that would be going? Because that is how this discussion is going.
 

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
BG, you keep saying using only even strength is illogical but you’ve been given reasons why people exclude it. You may disagree but excluding PP and PK isn’t being done without thought.

Okay, so why is all 4v4 play exuded as well?
 

SteveCangialosi123

Registered User
Feb 17, 2012
27,831
47,628
NJ
BG, you keep saying using only even strength is illogical but you’ve been given reasons why people exclude it. You may disagree but excluding PP and PK isn’t being done without thought.

Imagine if I kept ranting it’s illogical to not include all special team goals in a players +/- because you’re not accurately reflecting the goals that happen while a player is on the ice. People explain to me why they’re excluded but I just ignore their explanations and reiterate why I think it’s illogical. Can you see the illogic of how a discussion like that would be going? Because that is how this discussion is going.
There's no point, man. It's like talking to a wall.
 

Devils731

Registered User
Jun 23, 2008
12,199
16,108
Okay, so why is all 4v4 play exuded as well?

If you can give me all non-OT even strength data, it wouldn’t surprise me if that was better data than only 5 on 5 data. If your Even strength data includes OT though, then 5 on 5 only will be better.

The 5 on 5 data is probably such a majority of the time I doubt it would affect things much though.
 

devilsblood

Registered User
Mar 10, 2010
29,442
11,701
It's not really "interesting", it's metaphysically logical.

Either the metric works as a predictive value for team success, or it does not work as a predictive value for team success.

To exclude special teams from the broader data set is completely illogical.

The implied logic behind a Team's Corsi rating doesn't go away simply because you're on the PP or PK, it still applies. To argue otherwise, ironically means you do not believe in the core logic behind Corsi.

The fact that PP and PK time are "much smaller" data sets as people are noting is contextually completely irrelevant. Taken together, they BOLSTER your net available data, even if combined they may only represent something like 15% of the total data.
It's logical(not sure why it's metaphysical) if we accept the premises, but I think there must be an error in one of those premises.

Because, again, 5v5 corsi act 's as a very good predictor of success, pretty much 80% of top 10 teams make playoffs and 80% of bottom 10 teams do not, while overall corsi does not?

I'm just not sure which the weak premise is.

Edit: as noted below, doesn't make sense to include them all in one given that special team shots tend to go in the net at a higher rate. So obvious I looked right past it.

But I agree specialty team shots should be included in the equation some where, but they should be a seperate value.
 
Last edited:

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
This might be the essay that claims goal based stats are better for predictions then shot based stats.

Goal-based Metrics Better Than Shot-based Metrics at Predicting Hockey Success

It is dense as **** though. Totally beyond me.

If you notice, his models "all-in" pooled data sets outperformed the 5v5 data sets. Not shocking to me, but I thought I'd just point that out given the very recent discussion.

And...... Hmmmmmm........

Given the poor value of shot-based metrics at predicting team success, it was then not unexpected that shot-based metrics were also poor measures of individual player contributions to team success.

I'm going to print this out and enjoy reading it later.
 

Devils731

Registered User
Jun 23, 2008
12,199
16,108
This might be the essay that claims goal based stats are better for predictions then shot based stats.

Goal-based Metrics Better Than Shot-based Metrics at Predicting Hockey Success

It is dense as **** though. Totally beyond me.

I just got to glance through it real quick but it appears he’s taking the end of the season and saying goal differential explains standings better than shot differential does. I don’t think he will find anyone arguing against that. It seemed a lot of writing to confirm what was common thought, not a bad thing.

But if I was examining teams halfway through a season, would I see teams with good goal differentials and bad shot differentials keep their good goal differential all season or would they start drifting towards what their bad shot differential suggests. I believe it’s the latter but I haven’t run any analysis.
 

Devils731

Registered User
Jun 23, 2008
12,199
16,108
It's logical(not sure why it's metaphysical) if we accept the premises, but I think there must be an error in one of those premises.

Because, again, 5v5 corsi act 's as a very good predictor of success, pretty much 80% of top 10 teams make playoffs and 80% of bottom 10 teams do not, while overall corsi does not?

I'm just not sure which the weak premise is.

One of Corsi’s tenets is that teams will drift towards a normal team shooting percentage over an entire season. PP and PK vary wildly, even across an entire season, so shot based metrics for that don’t work as well.
 

BenedictGomez

Corsi is GROSSLY overrated
Oct 11, 2007
40,436
7,745
PRNJ
I just got to glance through it real quick but it appears he’s taking the end of the season and saying goal differential explains standings better than shot differential does. I don’t think he will find anyone arguing against that. It seemed a lot of writing to confirm what was common thought, not a bad thing.

But if I was examining teams halfway through a season, would I see teams with good goal differentials and bad shot differentials keep their good goal differential all season or would they start drifting towards what their bad shot differential suggests. I believe it’s the latter but I haven’t run any analysis.

Wrong.

He conducted a linear regression analysis.

What this means is that the results are exactly the opposite of what you're suggesting. Namely, that yes, the GD held up and was more accurate in predicting future success than the Corsi or other similar but tweaked Shot counting models.
 

Ad

Upcoming events

Ad

Ad

-->