Fancy Stats

[hobarth]

Besides the recent contract extension signed by Gallagher being less compared to Mitch Marner's contract, he's put up a lot less points than him. So wouldn't that mean Marner is better because of the offense he can produce?

Not sure what you're on about here, I was comparing a small player with a big heart against big players as a means to show that smaller players can be tough to play against. I wouldn't consider Marner big.
I think the closest we have to Gallagher is maybe Simmonds who at 6'2" isn't short but he weighs 185 lbs, Gallagher is 5'9". 184 lbs.

 

[Menzinger]

But Corsi wasn't wrong, was it? Carlyle's run and gun luck ran out, just like every sane person was predicting. In fact, once it ran out he set the NHL all time record for the most shots against in a season, both total and per game (36+ if I remember).

Not that Corsi is the be all end all, it's a very basic counting stat just slightly better than +/-, but that seems like an odd thing to pick at when the results ended up exactly where they were predicted to.

Good post.

Folks used to defend Caryle by bringing up thr shot quality over quantity debate, but his teams were getting shelled from everywhere including high danger areas. Stats said the 2013 team overperformed and got lucky it was a shoterned season, and that seemed to be the case too based on how that team looked the year after thag

 

[LeafsNation75]

Not sure what you're on about here, I was comparing a small player with a big heart against big players as a means to show that smaller players can be tough to play against. I wouldn't consider Marner big.
I think the closest we have to Gallagher is maybe Simmonds who at 6'2" isn't short but he weighs 185 lbs, Gallagher is 5'9". 184 lbs.

I'm saying if people want to compare Marner with Gallagher the only negative for Marner would be his contract, where as Gallagher just signed a 6 year extension worth $6.5 million AAV starting in the 2021-22 season.

However besides that Marner has proven to be a much better player.

 

[hobarth]

Good post.

Folks used to defend Caryle by bringing up thr shot quality over quantity debate, but his teams were getting shelled from everywhere including high danger areas. Stats said the 2013 team overperformed and got lucky it was a shoterned season, and that seemed to be the case too based on how that team looked the year after thag

A few years back Colorado finished on top of the Central with horrendous Corsi, we see streaks where bad teams win, not the corsi battle, and games where the better team loses but wins all the fancy stats.

The only true stat is goals, the team that scores more wins, every time. Plus/minus measures the # of goals that are scored when a player is on the ice, rarely if ever does a poor corsi player end up with plus stats, fancy stats are another way of expressing plus/minus.

 

[Dekes For Days]

A few years back Colorado finished on top of the Central with horrendous Corsi

By riding an unsustainable shooting percentage, and some of the best goaltending in the league. They then lost in the first round (to Minnesota...), and dropped to last in the division the following year. In fact, they were the 28th ranked team in the league over the subsequent 3 years. Not a great example.

Plus/minus measures the # of goals that are scored when a player is on the ice

Actually it doesn't. It only counts certain situations.

fancy stats are another way of expressing plus/minus.

"Advanced statistics" offer much, much, much better ways of expressing what plus/minus is badly trying to show.

 

[Bomber0104]

Not a pro or con comment about advanced stats, but the high/low danger shot thing has always been interesting. Just a few inches difference could mean the difference between high danger zone and a less dangerous zone. Seems like those stats could be error prone.

And this is precisely the problem with those particular stats.

Of course the problem could be solved instantly if the farcical concept of "Danger" was dropped (a wild over-reach), and just present the statistic for what it actually is measuring: distance from the net.

All shots can be measured based on their distance and then an average taken. But again, there's no measure of shot quality with that, is there?

So because that stat is a bit boring and doesn't really tell you a whole lot, what these blog stats people did was arbitrarily draw up "zones" on the ice and then give them an arbitrarily made-up name (High/Medium/Low).

Sounds cool, but what they're actually doing is pretty lazy/flawed and would probably get you a B- in a high school stats class.

 

[ACC1224]

A few years back Colorado finished on top of the Central with horrendous Corsi, we see streaks where bad teams win, not the corsi battle, and games where the better team loses but wins all the fancy stats.

The only true stat is goals, the team that scores more wins, every time. Plus/minus measures the # of goals that are scored when a player is on the ice, rarely if ever does a poor corsi player end up with plus stats, fancy stats are another way of expressing plus/minus.

Colorado was just lucky that year.

 

[Big Muddy]

Colorado was just lucky that year.

LOL. Don't forget the goaltending.

 

[LeafsNation75]

Colorado was just lucky that year.

Are you referring to the 2013-14 season when they had 112 points and Nathan MacKinnon was a rookie? It was only three years later when they finished last place overall in 2016-17 with just 48 points.

 

[Zybalto]

A few years back Colorado finished on top of the Central with horrendous Corsi, we see streaks where bad teams win, not the corsi battle, and games where the better team loses but wins all the fancy stats.

The only true stat is goals, the team that scores more wins, every time. Plus/minus measures the # of goals that are scored when a player is on the ice, rarely if ever does a poor corsi player end up with plus stats, fancy stats are another way of expressing plus/minus.

On the Leafs D this year:

Best 5v5 Corsi (minimum 300 minutes played):

1. Sandin: 54.76
2. Muzzin: 54.05
3. Barrie: 53.15
4. Holl: 52.14
5. Rielly: 51.91
6. Marincin: 50.87
7. Ceci: 50.70
8. Dermott: 50.41

Top +/- from that list:

1. Dermott: +14
2. Holl: +13
3. Muzzin: +12
4. Ceci: +7
5. Rielly: +6
6. Marincin: -1
7. Sandin: -7
8. Barrie: -7

Quite a bit of fluctuation in the rankings between the two. With goaltending as erratic as it was, not really a surprise I guess. I think both stats need to take into account usage and quality of competition/teammates among other things too.

In the Colorado case, a top 5 powerplay, top 5 goaltending and the 2nd ranked PK can take you pretty far. Like +/-, Corsi is used 5v5 and wont take account their insane special teams and goalkeeping.

They were an Abysmal 5v5 team though and lost out in the first round that season (dead last for team corsi in the playoffs). Their powerplay and goalkeeping fell off a bit and thats all it took against wild card qualifier Minnesota.

 

[Dekes For Days]

just present the statistic for what it actually is measuring: distance from the net.

But that's not all it's measuring. That's one part of what it's measuring. You're essentially suggesting we make the stat worse for no reason, just because you don't understand it.

what these blog stats people did was arbitrarily draw up "zones" on the ice and then give them an arbitrarily made-up name

None of the things you claim to be arbitrary are arbitrary.

 

[ACC1224]

On the Leafs D this year:

Best 5v5 Corsi (minimum 300 minutes played):

1. Sandin: 54.76
2. Muzzin: 54.05
3. Barrie: 53.15
4. Holl: 52.14
5. Rielly: 51.91
6. Marincin: 50.87
7. Ceci: 50.70
8. Dermott: 50.41

Top +/- from that list:

1. Dermott: +14
2. Holl: +13
3. Muzzin: +12
4. Ceci: +7
5. Rielly: +6
6. Marincin: -1
7. Sandin: -7
8. Barrie: -7

Quite a bit of fluctuation in the rankings between the two. With goaltending as erratic as it was, not really a surprise I guess. I think both stats need to take into account usage and quality of competition/teammates among other things too.

In the Colorado case, a top 5 powerplay, top 5 goaltending and the 2nd ranked PK can take you pretty far. Like +/-, Corsi is used 5v5 and wont take account their insane special teams and goalkeeping.

They were an Abysmal 5v5 team though and lost out in the first round that season (dead last for team corsi in the playoffs). Their powerplay and goalkeeping fell off a bit and thats all it took against wild card qualifier Minnesota.

What would be considered a good Corsi number?

 

[Notsince67]

A few years back Colorado finished on top of the Central with horrendous Corsi, we see streaks where bad teams win, not the corsi battle, and games where the better team loses but wins all the fancy stats.

The only true stat is goals, the team that scores more wins, every time. Plus/minus measures the # of goals that are scored when a player is on the ice, rarely if ever does a poor corsi player end up with plus stats, fancy stats are another way of expressing plus/minus.

Games a bit more complicated than that

 

[Notsince67]

What would be considered a good Corsi number?

Id say 100 is good

 

[ACC1224]

Id say 100 is good

So Sandin at 54 isn’t very good?

 

[Zybalto]

What would be considered a good Corsi number?

It's pretty straightforward when you are doing team rankings in that a score above 50 means, as a team, you are directing more shots towards the net than your opponents and we can directly compare team corsi against the rest of the league.

When it comes to players, I think Corsi REL numbers are the best way to compare as your team may be a good corsi team (as the Leafs are) and comparing against how the players does against his peers with the same goalie, structure, etc.. is more accurate to see his impact.

I'm not a massive fan of Corsi myself (xGF% is far superior so why not use that?) but it still gives a decent look at players that were able to drive posession.

Out of the 164 NHL forwards with at least 800 minutes played last year, the top Corsi REL player was Pettersson and the worst was Kuznetsov. These two players were both the league MVP and a real anchor when it comes to driving possession.

 

[Zybalto]

So Sandin at 54 isn’t very good?

Sandin is a wonderful possession player but had the misfortune of being a casualty of the wave of D injuries in an indirect way as he really gelled with a more experienced stay at home guy like Ceci (this is when he got all that media buzz early in the year) but once Ceci went down with that high ankle sprain, and the Leafs were already short handed on D, he was forced to play with bad matches or guys that wernt quite ready (Liljegren)

Here are his 5v5 numbers with and away from Ceci:

With Ceci:

Ozone faceoff%: 42
CF%: 55.78 (The highest number of any Leafs pairing with 50+ minutes played together last season)
xGF%: 59.17 (The highest number of any Leafs pairing with 50+ minutes played together last season)


Away from Ceci:

Ozone faceoff%: 63
CF%: 54.36
xGF%: 45.36

With Ceci, he was able to push the play knowing he had a guy behind him he could trust to cover any errors and they played some pretty defensive minutes but even with being uber sheltered after Ceci went down, he could never find the same level of success with other partners and so he scratched once the team got healthy enough.

 

[LeafsNation75]

A few years back Colorado finished on top of the Central with horrendous Corsi, we see streaks where bad teams win, not the corsi battle, and games where the better team loses but wins all the fancy stats.

The only true stat is goals, the team that scores more wins, every time. Plus/minus measures the # of goals that are scored when a player is on the ice, rarely if ever does a poor corsi player end up with plus stats, fancy stats are another way of expressing plus/minus.

That's why I never got into understanding how something like Corsi works. So I prefer looking at actual stats like you mentioned.

 

[Dekes For Days]

That's why I never got into understanding how something like Corsi works. So I prefer looking at actual stats like you mentioned.

Corsi is an "actual stat".

 

[Nineteen67]

On the Leafs D this year:

Best 5v5 Corsi (minimum 300 minutes played):

1. Sandin: 54.76
2. Muzzin: 54.05
3. Barrie: 53.15
4. Holl: 52.14
5. Rielly: 51.91
6. Marincin: 50.87
7. Ceci: 50.70
8. Dermott: 50.41

Top +/- from that list:

1. Dermott: +14
2. Holl: +13
3. Muzzin: +12
4. Ceci: +7
5. Rielly: +6
6. Marincin: -1
7. Sandin: -7
8. Barrie: -7

Quite a bit of fluctuation in the rankings between the two. With goaltending as erratic as it was, not really a surprise I guess. I think both stats need to take into account usage and quality of competition/teammates among other things too.

In the Colorado case, a top 5 powerplay, top 5 goaltending and the 2nd ranked PK can take you pretty far. Like +/-, Corsi is used 5v5 and wont take account their insane special teams and goalkeeping.

They were an Abysmal 5v5 team though and lost out in the first round that season (dead last for team corsi in the playoffs). Their powerplay and goalkeeping fell off a bit and thats all it took against wild card qualifier Minnesota.

What is the Corsi for each player in the games Sandin played?

 

[LeafsNation75]

Corsi is an "actual stat".

I realize it's a stat for those who follow it and understand how it works.

However like I said before it's something I haven't bothered to learn how it works, which is why I like looking at what a player does in a game with their goals, assists, points, and plus/minus.

 

[Bomber0104]

But that's not all it's measuring. That's one part of what it's measuring. You're essentially suggesting we make the stat worse for no reason, just because you don't understand it.

That's almost entirely what it's measuring, with no measure of quality given.

A shot from David Steckel is given the almost same value as a shot from Alexander Ovechkin, taken from the same place.

There's no measure of the actual talent of the player given to the model.

And this is why these stats are seen as a joke except from people like you.

None of the things you claim to be arbitrary are arbitrary.

Giving things names like "High", "Medium", and "Low" are as arbitrary as anything.

It means nothing.

You might get a B- in a high-school stats class with this kind of stuff ,but in a serious way these ways of modelling and presenting data don't fly in the real world.

I invite our unbiased moderator @Menzinger to address these issues, since he "likes" your post as well

 

[Bomber0104]

Corsi is an "actual stat".

It really doesn't mean anything though.

That's why it doesn't correlate to wins, playoff wins, Stanley Cup winners, etc.

It's a stat, surely.

But it's not one that is significant in any way.

CORSI/Fenwick were once used to suggest Mikhail Grabovski and Clarke Macarthur were some of the top players in the ENTIRE NHL here on this board.

How could people possibly have believed that watching the games?!

 

[Bomber0104]

Sandin is a wonderful possession player but had the misfortune of being a casualty of the wave of D injuries in an indirect way as he really gelled with a more experienced stay at home guy like Ceci (this is when he got all that media buzz early in the year) but once Ceci went down with that high ankle sprain, and the Leafs were already short handed on D, he was forced to play with bad matches or guys that wernt quite ready (Liljegren)

Here are his 5v5 numbers with and away from Ceci:

With Ceci:

Ozone faceoff%: 42
CF%: 55.78 (The highest number of any Leafs pairing with 50+ minutes played together last season)
xGF%: 59.17 (The highest number of any Leafs pairing with 50+ minutes played together last season)


Away from Ceci:

Ozone faceoff%: 63
CF%: 54.36
xGF%: 45.36

With Ceci, he was able to push the play knowing he had a guy behind him he could trust to cover any errors and they played some pretty defensive minutes but even with being uber sheltered after Ceci went down, he could never find the same level of success with other partners and so he scratched once the team got healthy enough.

Sandin looked categorically outmatched by way better opposition down low and got his face plastered into the boards the few times he saw ice this season, but apparently he was good?

It sounds like your stats should be taken very, very seriously!

Amazing the things that show up on blogs vs actually watching what happened!

 

[hobarth]

Are you referring to the 2013-14 season when they had 112 points and Nathan MacKinnon was a rookie? It was only three years later when they finished last place overall in 2016-17 with just 48 points.

That sounds about right, it seems odd that, I thought the next year that Colorado finished so poorly, tanked maybe, they could be so good and then so bad for entire years in both cases. A whole pile of things make for good stats, Komisarek was a great signing by Burke except TO didn't have a Markov clone to prop him up.

On the Leafs D this year:

Best 5v5 Corsi (minimum 300 minutes played):

1. Sandin: 54.76
2. Muzzin: 54.05
3. Barrie: 53.15
4. Holl: 52.14
5. Rielly: 51.91
6. Marincin: 50.87
7. Ceci: 50.70
8. Dermott: 50.41

Top +/- from that list:

1. Dermott: +14
2. Holl: +13
3. Muzzin: +12
4. Ceci: +7
5. Rielly: +6
6. Marincin: -1
7. Sandin: -7
8. Barrie: -7

Quite a bit of fluctuation in the rankings between the two. With goaltending as erratic as it was, not really a surprise I guess. I think both stats need to take into account usage and quality of competition/teammates among other things too.

In the Colorado case, a top 5 powerplay, top 5 goaltending and the 2nd ranked PK can take you pretty far. Like +/-, Corsi is used 5v5 and wont take account their insane special teams and goalkeeping.

They were an Abysmal 5v5 team though and lost out in the first round that season (dead last for team corsi in the playoffs). Their powerplay and goalkeeping fell off a bit and thats all it took against wild card qualifier Minnesota.

It looks like Corsi has Sandin, Muzzin and Barrie as all-star d-men meanwhile +/- has Dermott, Holl and Muzzin as TO's top d-men, I don't agree with either conclusion except I do think Muzzin should be in the top 3 for TO in both ratings. My conclusion and I think it's amply supported by both methods of rating is that both are flawed and neither is superior to the other.

In 2013/14 Colorado finished with 112 points but their team corsi was horrible, an anomaly, and that team continued with it's horrible corsi putting up 90 and 82 points until finally bottoming out with 48 points but in 2013/2014 they lost to Minny because of their corsi? In 2018/19 Tampa put together a near historic season, they won nearly every game, certainly every fancy stat rating but then lost to Columbus in the playoffs. I don't know what conclusion can be derived from these 2 teams but corsi itself doesn't explain it and neither does +/-.