Corsi, shot quality, and the Toronto Maple Leafs

Cap'n Flavour

Registered User
Mar 8, 2004
4,938
1,618
Flavour Country
Who is claiming that advanced stats will predict the results of 8 game stretches anyways?

Corsi/Fenwick don't even have a strong correlation with goal differential over the course of a full season, it's something like 0.3. Special teams, goaltending (sv%), sh%, and luck have a major role in scoring goals and winning game too. Every time I read some half-baked criticism of advanced stats this is pointed out as if it's a fatal flaw of possession statistics. Newsflash, the people with a clue working on this stuff already realized all of this. Possession is still correlated with winning, though, so it's blindingly obvious that you want to have good possession to have a better chance of winning games without running on luck.

What people like Mirtle were saying is that the Leafs' play early on in the season was unsustainable. It was. They said that the Leafs' good record even through February was masking some quite poor on-ice play and was going to backfire sooner or later - maybe not enough for the Leafs to miss the playoffs entirely, but possibly during the playoffs themselves. That was and is also true. The point about the Leafs' abysmal ROW and dependence on shootout points to balance poor regulation+OT play also stands. Did Corsi predict that the Leafs would lose eight pathetic games in a row? No, but so what? It never claimed to. Plus every time I read ridiculous statements like that, I imagine that clueless blowhard Cox blaming Jim Corsi for personally predicting the result of every game, sometimes incorrectly.
 
Last edited by a moderator:

Master_Of_Districts

Registered User
Apr 9, 2007
1,744
4
Black Ruthenia
Who is claiming that advanced stats will predict the results of 8 game stretches anyways?

Corsi/Fenwick don't even have a strong correlation with goal differential over the course of a full season, it's something like 0.3. Special teams, goaltending (sv%), sh%, and luck have a major role in scoring goals and winning game too. Every time I read some half-baked criticism of advanced stats this is pointed out as if it's a fatal flaw of possession statistics. Newsflash, the people with a clue working on this stuff already realized all of this. Possession is still correlated with winning, though, so it's blindingly obvious that you want to have good possession to have a better chance of winning games without running on luck.

What people like Mirtle were saying is that the Leafs' play early on in the season was unsustainable. It was. They said that the Leafs' good record even through February was masking some quite poor on-ice play and was going to backfire sooner or later - maybe not enough for the Leafs to miss the playoffs entirely, but possibly during the playoffs themselves. That was and is also true. The point about the Leafs' abysmal ROW and dependence on shootout points to balance poor regulation+OT play also stands. Did Corsi predict that the Leafs would lose eight pathetic games in a row? No, but so what? It never claimed to. Plus every time I read ridiculous statements like that, I imagine that clueless blowhard Cox blaming Jim Corsi for personally predicting the result of every game, sometimes incorrectly.

Not that it detracts from your argument (it assists it, if anything), but the correlation between possession stats and goal differential has been about 0.5 - 0.7 in the seasons for which data is available.

Just checked - r=0.57 for fenwick and goal ratio for the seasons of 2007-08 to 2011-12, inclusive.

And, of course, I know the temptation of some (not the poster I'm quoting, obviously, but others) will be to look at that and say: "r=0.57! That's only an r^2 of 0.32! What a terrible correlation! Fenwick is worthless. But if you correct for attenuation - which can be done if you know the reliability co-efficients for each variable over the sample in question (about 0.95 for fenwick and 0.55 for goal ratio, in this case) - it turns out that the true correlation between fenwick and goal ratio is about 0.79.

Not so terrible, I would say.
 
Last edited:

hatterson

Registered User
Apr 12, 2010
34,951
12,226
North Tonawanda, NY
Who is claiming that advanced stats will predict the results of 8 game stretches anyways?

In fairness, several of the recent posts in the thread were actually implying that, hence my reaction. Using the Leafs 8 game losing streak/collapse and (probable) miss of the playoffs as an "I told you so" isn't statistically honest since it relies on something just as anomalous as the previous good luck.

It's like me planning on flipping a fair coin 82 times and after the first 65 I have 40 heads and 25 tails. Obviously me claiming that I should finish with a "record" around 50 heads-32tails (~61% heads) is incorrect and people should be saying "That's unsustainable, it's very statistically unlikely that you keep getting that many heads". They would be right. But then if I go on a streak of 8 straight tails it doesn't suddenly prove them right since 8 straight tails is also a statistical anomaly.

"The Leafs play isn't sustainable" - I completely agree (and have argued as much during this thread). However simply because it wasn't sustainable did not predict an 8 game collapse and missing of the playoffs which means that using these things as an "I told you so" is being dishonest from a statistical sense.

On March 15th the Leafs had a record of 36-24-8, good for 80 points through 68 games, a 3rd place position in the conference, a 6 point gap on 9th, and a 94-95 point pace. Obviously that pace was slightly higher than you'd expect them to finish with given that it included some good luck from earlier in the season. However, even if we assume that they had "normal" luck since then and corsi tracked with results you'd expect a finish in the last 14 games of around 5-7-2 or so, maybe 4-7-3 or 4-8-2. Through the next 8 that might translate to something along the lines of 3-5 or 3-4-1 or 2-4-2. Those records right now would put them into the playoffs with a 3-6 point cushion (depending on who they came against) and a virtual lock at the spot.

My original reply was to the comment of "If the collapse of the 13-14 Leafs doesn't convince you that "advanced" stats are meaningful then nothing will." And was centered around the fact that advanced stats didn't predict a collapse like this and thus the collapse shouldn't convince you of anything.

What people like Mirtle were saying is that the Leafs' play early on in the season was unsustainable. It was. They said that the Leafs' good record even through February was masking some quite poor on-ice play and was going to backfire sooner or later - maybe not enough for the Leafs to miss the playoffs entirely, but possibly during the playoffs themselves. That was and is also true. The point about the Leafs' abysmal ROW and dependence on shootout points to balance poor regulation+OT play also stands. Did Corsi predict that the Leafs would lose eight pathetic games in a row? No, but so what? It never claimed to. Plus every time I read ridiculous statements like that, I imagine that clueless blowhard Cox blaming Jim Corsi for personally predicting the result of every game, sometimes incorrectly.

Actually their ROW has come back in line recently. They're 2 ahead of Washington and 3 behind Detroit while being 1 and 4 points behind respectively. Now they're no more out of line that Tampa, Detroit, Washington, San Jose, Minnesota, Phoenix and Vancouver. Their earlier shootout success has been offset mostly by the fact that they haven't been involved in a shootout since Jan 15th.
 

Diatomic

Mitch Matthewlander
Mar 12, 2013
9,178
81
Air Canada Centre
LvSvOvC-10GPCF_.png


Were just as bad as the worst teams in the NHL.
 

BraveCanadian

Registered User
Jun 30, 2010
14,519
3,357
Favorite phrase: "The Leafs, whose playoff hopes effectively ended this weekend with a wet fart of a loss to Winnipeg..."

No matter the numbers involved.. man what a gutless crew of players on the Leafs this year. No heart or drive at all during long stretches of games.

The article is just analytics fist bumping, though. How many times this season were they vindicated before the Leafs roared away on another good streak? 3 or 4 times? I've lost count.

And now it is "The Leafs won because of unbelievable luck. They lost because they're bad at NHL-level hockey."

While I agree about that last part.. they were "unbelievably" lucky for a long, long time. They were also unbelievably unlucky (and playing terribly too) to lose 8 straight without even a loser point. (Goaltending coming back down to earth while the team played even worse than normal in front of them imo)

Obviously there is a lot more to unravel.
 
Last edited:

Master_Of_Districts

Registered User
Apr 9, 2007
1,744
4
Black Ruthenia
While I agree about that last part.. they were "unbelievably" lucky for a long, long time. They were also unbelievably unlucky (and playing terribly too) to lose 8 straight without even a loser point. (Goaltending coming back down to earth while the team played even worse than normal in front of them imo)

Obviously there is a lot more to unravel.

Yeah - it's almost as if...randomness had a strong effect on results over a small segment of games.

Weird.
 

BraveCanadian

Registered User
Jun 30, 2010
14,519
3,357
Yeah - it's almost as if...randomness had a strong effect on results over a small segment of games.

Weird.

Yes, randomness when they lose over a handful of games but unbelievable luck when they buck the expectation over the better part of what.. 100 games?

Now imagine that if they had won few of those 8 games.

Suddenly they have continued to prove the inability of the current "advanced stats" models to predict anything wrong over a full season and half.

Back to the drawing board.
 

BraveCanadian

Registered User
Jun 30, 2010
14,519
3,357
Among other things, no one who should be taken seriously is using Corsi in isolation.

I'll be the first to admit that I am not up on the latest.

What is the latest?

Does the NHL or anyone else track possession time rather than using a proxy like Corsi?

I'd have to assume that the teams themselves track something like that even if it isn't publicly available.
 

Canadiens1958

Registered User
Nov 30, 2007
20,020
2,773
Lake Memphremagog, QC.
Teams

I'll be the first to admit that I am not up on the latest.

What is the latest?

Does the NHL or anyone else track possession time rather than using a proxy like Corsi?

I'd have to assume that the teams themselves track something like that even if it isn't publicly available.

NHL teams track possession times, zone times, etc BUT as a function of game plan objectives.

Since teams are not in the habit of making game plans public.....
 

Master_Of_Districts

Registered User
Apr 9, 2007
1,744
4
Black Ruthenia
Yes, randomness when they lose over a handful of games but unbelievable luck when they buck the expectation over the better part of what.. 100 games?

I'm not sure what you're even trying to say here - the phasing is unintelligible.

Now imagine that if they had won few of those 8 games.

Suddenly they have continued to prove the inability of the current "advanced stats" models to predict anything wrong over a full season and half.

It would prove nothing of the sort.

Although the fact that you think otherwise is amusing, to say the least.

Back to the drawing board.

Speak for yourself.
 

MVP of West Hollywd

Registered User
Oct 28, 2008
3,489
950
Being fair, the Leafs goal differential sucked for their record even before the streak. Since advanced stats community should care about goal differential more than wins, they were already right about Leafs regression. To an extent. Overall the Leafs still outperformed Corsi at ES this yr almost as much as anyone. Finishing 20th in ES with 30 Corsi with Edm and Buf being truly horrible compared to 27 and up, is still not lining up with possession. The Leafs even now are as much a possession to performance outlier as Anaheim, Colorado, New Jersey, Florida.
 
Last edited:

Bear of Bad News

Your Third or Fourth Favorite HFBoards Admin
Sep 27, 2005
13,345
26,516
Globe and Mail's Dave Shoalts on what he hates:

“Most hockey analytics geeks. Aside from having no sense of humour, they all act like they are the true sages of hockey simply because they came up with a few equations to state the obvious. The team that has the puck most usually wins. No kidding, Sherlock. That’s been true since Lord Stanley was talked into spending 35 bucks to buy a certain cup.â€

Source: http://www.torontosun.com/2014/05/03/sharks-movement-to-keep-mclellan-could-stall-coaching-carousel

Someone needs to buy Dave some smaller brushes to paint with. And it's the "geeks" who don't have a sense of humor?
 

JoelWarlord

Ex-Noob616
May 7, 2012
6,078
9,236
Halifax
The point about 3 of the top 10 Corsi teams being eliminated in the first round is interesting given that all 3 played other top 10 Corsi teams.
 

Ad

Upcoming events

Ad

Ad

-->