Jets Advanced Stats thread

[Guffman]

Some of these responses are hilarious:

The Guffman Index flawlessly predicts all 16 playoff teams... again!

I just wanted to chime in here to say The Guffman stat remains THE most notable stat in terms of team success. It was yet another year where all 16 playoff teams also had registered the top 16 stats in The Guffman index. I discovered this stat back in March, 2019 and posted it on this subforum...

forums.hfboards.com

You guys need to step up your game.

 

[DRW204]

Imagine all the “advanced stats” posters never mentioning goal differential at all but droning on and and on about Corsi.

Well I don't think it's considered an advanced stat so maybe that's why. More on the traditional or simple side. By-products of goal differential, or its components (GF or GA) get mentioned quite a bit though. But hey, congrats your discovery.

 

[gojetsgo]

 

[BobTheSolarsystem]

Holy neurotic... let's hope this personality stays for the playoffs!

 

[DRW204]

not really advanced stats.... but just looking at some team goal scoring trends and how the %'s and totals lay over the past couple years (sorted by highest-GF teams)

2021-2022:
FLA, STL, COL even CBJ are examples of good depth and lower concentration on the top-end in goal-scoring.
TOR had fantastic top-end as well as real good depth.
DAL had strong top end, poor depth.

22-23
EDM real strong top-end, and good dpeth
SEA, VGK, CAR, STK not-so-good top-end, and good depth

 

[DRW204]

By xgf% (5v5) the final four composed of the

1st, 5th, 10th and 16th best teams in the regular season (NST) or 1st, 7th, 10th and 12th (MP)


By all strengths goal diff:
4th, 7th, 9th and 16th.

 

[DRW204]

not really advanced stats, but just a summary of some draft stats.

so looking at players drafted mid-to-late 1st, so 15th-31st (i know this was considered 1st pick of the 2nd for many years but is a 1st now! lol) from 2005-2019. concentrating on Fwds so looking at PPG.

there were 157 Fwds drafted from 2005-2019 b/w picks 15-31. their per-82 gp stats as of today is categorized as follows (& of course this will change every season)

5 of the fwds in the 0 tier are from the 2018 or 2019 drafts, so there is still a possibility of them doing something. not so sure about anything from 6+ years ago

going down to a minimum of 82 career gp (1 season) - out of 108 players:

again stats as of today, this will change over time.

 

[DRW204]

 

[lomiller1]

Imagine all the “advanced stats” posters never mentioning goal differential at all but droning on and and on about Corsi.

Predictive vs descriptive.

Cosri predicts goal deferential better than pas goal differential does, so if you want to know about future goal differential Corsi is a better stat than past goal differential. If all you care about is games that have already been played then you can ignore goal differential and go straight to wins and losses.

In neither case is goal differential all that useful.

 

[Buffdog]

Carolina is 3-3

Boston is undefeated

 

[Guffman]

Predictive vs descriptive.

Cosri predicts goal deferential better than pas goal differential does, so if you want to know about future goal differential Corsi is a better stat than past goal differential. If all you care about is games that have already been played then you can ignore goal differential and go straight to wins and losses.

In neither case is goal differential all that useful.

Cool. After 20 games, I want you to make predictions for the rest of the season based on Corsi and we’ll see how that turns out.

 

[lomiller1]

Cool. After 20 games, I want you to make predictions for the rest of the season based on Corsi and we’ll see how that turns out.

It's been proven to turn out better than predictions made on goal differential...

 

[Guffman]

It's been proven to turn out better than predictions made on goal differential...

OK, lets say a team plays 40 games and accumulated a 53% Corsi over the season. Let’s also say that in the past 10 games, their Corsi was at 45%. What’s your prediction of future performance and explain your underlying logic.

I assume you can do this. Looking forward to your explanation.

 

[DRW204]

I'm looking forward to the predictions challenge

Wondering if you looked at last season's Corsi and goal differential (assuming all strengths) rankings at the 20 gp mark for each team, how did those teams perform rest of the 62 gp

 

[Romang67]

Most predictive statistics rely in some way or another on the law of large numbers. There aren't enough goals scored for or against a team in a season for regression to be meaningful.

The predictive power (or lack thereof) of Corsi and goal differential have different causes. Corsi is a high volume event, but may not be predictive because it doesn't actually indicate success in hockey. Goals isn't a high volume event, so before regression to the mean would be relevant, the team composition or opposition would likely already have changed.

For clarity, there also aren't enough Corsi events for the law of large numbers to be relevant. Not now, and not if you kept accumulating Corsi events until the concept of hockey was forgotten. But there are enough of them for regression to be meaningful over a season. Probably.

 

[Guffman]

I'm looking forward to the predictions challenge

Wondering if you looked at last season's Corsi and goal differential (assuming all strengths) rankings at the 20 gp mark for each team, how did those teams perform rest of the 62 gp

I would challenge that Garett poster with that each year with that particular challenge and he always passed.

Here is the thing with Corsi and it’s so-called “predictive element”. Sure, it makes sense that if you outshoot opponents over a period of time, maybe it is somewhat reasonable that you will continue to outshoot them in the future and the better your shot differential, perhaps the more relative goals you score.

BUT…

This is taking stats from individual players and individual teams and smashing it up into some big macro ball of meaningless crap. Why is it crap? Well, let’s say you have a sniper like Kyle Connor and his line fires off 8 shots and let’s say his line plays against a stone-handed defensive line that takes 10 shots… most miss, some are blocked and some are primitive with a low threat to score. Gee, KC and his mates suck out on Corsi and “predictively”, that doesn’t bode well for the Jets!

Of course, that’s all bullshit. I high skilled team with superior shooting percentage that pass up the easy shot for creative high skill plays with a greater chance at scoring will generally do poorly at Corsi.

Corsi doesn’t measure the quality of the shot. Corsi doesn’t factor in the quality of the player taking the shot. Corsi doesn’t even care if the shot misses the net. Why? Because it’s a completely lazy statistic that’s easy to compile without digging deeper. Proponents of Corsi say all those things doesn’t matter (surprisingly) because they have some sort of statistical proof from their big smushed up macro ball kinda has some predictive value… but you cannot actually apply it well at an individual team or player level because then those disgusting hairballs will dirty up this very simple statistic.

At the end of the year, the 16 playoff teams almost always have the 16 greatest goal differential and top Corsi teams are spread around between playoff and non-playoff teams. But then they scream “it’s a predictive statistic!!”. OK, if that’s the case, let’s pull out the Corsi after 20 or 40 games and make your predictions.

Any Corsi proponent willing to finally step up?

 

[DRW204]

Cool. After 20 games, I want you to make predictions for the rest of the season based on Corsi and we’ll see how that turns out.

I would challenge that Garett poster with that each year with that particular challenge and he always passed.

Here is the thing with Corsi and it’s so-called “predictive element”. Sure, it makes sense that if you outshoot opponents over a period of time, maybe it is somewhat reasonable that you will continue to outshoot them in the future and the better your shot differential, perhaps the more relative goals you score.

BUT…

This is taking stats from individual players and individual teams and smashing it up into some big macro ball of meaningless crap. Why is it crap? Well, let’s say you have a sniper like Kyle Connor and his line fires off 8 shots and let’s say his line plays against a stone-handed defensive line that takes 10 shots… most miss, some are blocked and some are primitive with a low threat to score. Gee, KC and his mates suck out on Corsi and “predictively”, that doesn’t bode well for the Jets!

Of course, that’s all bullshit. I high skilled team with superior shooting percentage that pass up the easy shot for creative high skill plays with a greater chance at scoring will generally do poorly at Corsi.

Corsi doesn’t measure the quality of the shot. Corsi doesn’t factor in the quality of the player taking the shot. Corsi doesn’t even care if the shot misses the net. Why? Because it’s a completely lazy statistic that’s easy to compile without digging deeper. Proponents of Corsi say all those things doesn’t matter (surprisingly) because they have some sort of statistical proof from their big smushed up macro ball kinda has some predictive value… but you cannot actually apply it well at an individual team or player level because then those disgusting hairballs will dirty up this very simple statistic.

At the end of the year, the 16 playoff teams almost always have the 16 greatest goal differential and top Corsi teams are spread around between playoff and non-playoff teams. But then they scream “it’s a predictive statistic!!”. OK, if that’s the case, let’s pull out the Corsi after 20 or 40 games and make your predictions.

Any Corsi proponent willing to finally step up?

well idk if im a corsi proponent you're talking about. i think it has it's uses but there's lots of other variables in the game (goaltending, finishing, special teams etc) that are important that some of the corsi or xGF% pumpers usually ignore so i guess i kinda agree with you. also when looking at games that happened and evaluating they seem to just look at just outshooting the opponent meanwhile Helle posted a sub 900 sv% and the team gave up 2 special team goals. but sure those shot attempts!

just looking at last season through 20 games

the top-16 teams in Corsi% were (in order)

CAR, NJD, FLA, CGY, NYR, BUF, OTT, LAK, BOS, TOR, WSH, SEA, TBL, PIT, COL, DAL,

the top-16 teams in goal-differential (GD) - assuming all-strengths (in order)

BOS, NJD, DAL, VGK, COL, SEA, NYI, WPG, NYR, TBL, PIT, BUF, TOR, DET, FLA, MIN

The final 62-GP these were the best teams in terms of points:

BOS, EDM, CAR, TOR, COL, MIN, NYR, LAK, DAL, VGK, NJD, BUF, TBL, SEA, NSH, CGY

so 12/16 based on corsi% remained as a top-16 team. largest swings were EDM as 19th placed Corsi (at the 20-GM mark) being the 2nd best team the final 62-gp. largest swing the other way was FLA 3rd best CF% team being the 18th best team the rest of the way.

now using GD: 11/16 teams who were the best in GD at the 20-gm mark were still a top-16 the rest of the 62-gp. the largest swing EDM again was the 19th best GD team through 20-gms and 2nd best team rest of the way. DET was t-12th best team in GD and then finished with the 26th most pts.

GD seemed to have a lesser average variance when comparing the leaders at the 20-gm mark to the final 62-gp across the entire league. When comparing just the 16 best teams from the final 62-gp CF% was a smidge better.

anyway, i think at the end of the day a combo of outshooting (CF%), outshooting (xGF%) and outscoring (GF%) is what you want. from the looks of the last few years looks like the worst CF% team to win the cup was Vegas, but most are in the top-10ish range.

 

[Whileee]

So far this season the Jets are getting a lot more shot attempts than their opponents, but they aren't getting many shots from the slot and centre of the ice. That's limited their scoring.

Defensively, they've actually been quite good at preventing both the volume and quality of shots against, 5v5.

So far, the Jets' main deficits are on special teams and in goaltending.

 

[surixon]

So far this season the Jets are getting a lot more shot attempts than their opponents, but they aren't getting many shots from the slot and centre of the ice. That's limited their scoring.

Defensively, they've actually been quite good at preventing both the volume and quality of shots against, 5v5.

So far, the Jets' main deficits are on special teams and in goaltending.

View attachment 760244

Pretty much the hallmarks of a Bowness team at 5 on 5.

Can't remember how his Stars teams did on special teams.

 

[Buffdog]

well idk if im a corsi proponent you're talking about. i think it has it's uses but there's lots of other variables in the game (goaltending, finishing, special teams etc) that are important that some of the corsi or xGF% pumpers usually ignore so i guess i kinda agree with you. also when looking at games that happened and evaluating they seem to just look at just outshooting the opponent meanwhile Helle posted a sub 900 sv% and the team gave up 2 special team goals. but sure those shot attempts!

just looking at last season through 20 games

the top-16 teams in Corsi% were (in order)

CAR, NJD, FLA, CGY, NYR, BUF, OTT, LAK, BOS, TOR, WSH, SEA, TBL, PIT, COL, DAL,

the top-16 teams in goal-differential (GD) - assuming all-strengths (in order)

BOS, NJD, DAL, VGK, COL, SEA, NYI, WPG, NYR, TBL, PIT, BUF, TOR, DET, FLA, MIN

The final 62-GP these were the best teams in terms of points:

BOS, EDM, CAR, TOR, COL, MIN, NYR, LAK, DAL, VGK, NJD, BUF, TBL, SEA, NSH, CGY

so 12/16 based on corsi% remained as a top-16 team. largest swings were EDM as 19th placed Corsi (at the 20-GM mark) being the 2nd best team the final 62-gp. largest swing the other way was FLA 3rd best CF% team being the 18th best team the rest of the way.

now using GD: 11/16 teams who were the best in GD at the 20-gm mark were still a top-16 the rest of the 62-gp. the largest swing EDM again was the 19th best GD team through 20-gms and 2nd best team rest of the way. DET was t-12th best team in GD and then finished with the 26th most pts.

GD seemed to have a lesser average variance when comparing the leaders at the 20-gm mark to the final 62-gp across the entire league. When comparing just the 16 best teams from the final 62-gp CF% was a smidge better.

View attachment 757741

anyway, i think at the end of the day a combo of outshooting (CF%), outshooting (xGF%) and outscoring (GF%) is what you want. from the looks of the last few years looks like the worst CF% team to win the cup was Vegas, but most are in the top-10ish range.

I think that at the end of the day, the teams that win the cup rank high in CF% because they're good teams

Where stats people go wrong is by flipping that and saying that teams are good BECAUSE they rank high in CF%

 

[Slimy Sculpin]

So far this season the Jets are getting a lot more shot attempts than their opponents, but they aren't getting many shots from the slot and centre of the ice. That's limited their scoring.

Defensively, they've actually been quite good at preventing both the volume and quality of shots against, 5v5.

So far, the Jets' main deficits are on special teams and in goaltending.

View attachment 760244

"main deficits are on special teams and in goaltending" I think the "eye test" confirms this.

 

[DRW204]

"main deficits are on special teams and in goaltending" I think the "eye test" confirms this.

i have faith the Goaltending will turn itself around with Helle which will bleed into a bit on the PK. he's already been playing better as of late.

PP they've made a switch w/ Perfetti up there, however, i think it's more of the style than the players that's the problem.

they've definitely improved on finishing a bit. but i am still in the "need to see more" camp on that side of things to say oh this team is great at scoring. im old enough to remember last year when some key players started out on torrid paces, got fans' expectations set super high, and then regressed back closer to their career norms and everyone was wondering what happened/disappointed.

 

[BoneDocUK]

i have faith the Goaltending will turn itself around with Helle which will bleed into a bit on the PK. he's already been playing better as of late.

PP they've made a switch w/ Perfetti up there, however, i think it's more of the style than the players that's the problem.

they've definitely improved on finishing a bit. but i am still in the "need to see more" camp on that side of things to say oh this team is great at scoring. im old enough to remember last year when some key players started out on torrid paces, got fans' expectations set super high, and then regressed back closer to their career norms and everyone was wondering what happened/disappointed.

I hear ya, but I'm still picking Apples to finally hit 30.




Or 10.

 

[Frank Cannon PI]

I hear ya, but I'm still picking Apples to finally hit 30.




Or 10.

So basically, Ehlers GPG pace in the playoffs.

 

[Guffman]

CF% and their so-called predictive value of future success

Let me understand the logic of this. Having a high historical CF% is predictive of future success because:

1. If you have a high CF% historically, it makes sense that you can continue to have a high CF% in the future
2. If you outshoot your opponents over a period of time, you will likely outscore your opponents over a period of time.
3. If you outscore your opponents over a period of time, you will likely have a higher victory % over time.

Is this the logic? If so, to prove out the predictive value of CF%, we should be going through each of these bullets rather than jump from having a high current CF% to future wins. The logic first needs to go from having a high current CF% to a high future CF%, right?

I may take a look at the CF% at the 20 game mark and see how well the logic holds but first, let's agree on the logic beforehand.