Wild Card Race (Western Conference)

[DJones]

http://bleacherreport.com/articles/2...ek-for-march-2

Not exactly related but I have never seen bleacher report give the Flames this much credit.

[bomber317]

Just adblock the chart already. Save yourself the headache. You knew there was an issue with the methodology when they had Dallas ahead of Calgary in percentage of making the playoffs a few weeks back. This was when Calgary was up by 5-7 points on Dallas.

Trolling chart @ best for Flames fans, since we all know our team doesn't play the style that is conductive for good corsi numbers. So why use a chart that uses corsi to predict our playoff chances?
Stick with the other site predictors like sportsclubstats.

[heep223]

Originally Posted by bomber317

You knew there was an issue with the methodology when they had Dallas ahead of Calgary in percentage of making the playoffs a few weeks back. This was when Calgary was up by 5-7 points on Dallas.

Seriously? Oh man there goes any credibility.

And here comes the retort that "just because it's wrong doesn't mean it's wrong" lol

[Resolute 14]

Originally Posted by CorsiHockeyLeague

This discussion is pretty similar to the one that came up earlier in the thread when that graph was posted, so no point going over old ground. All I'll say is it's based on a methodology with a record of making correct predictions about future team success. That doesn't mean it's always right and it doesn't mean the Flames won't make the playoffs in spite of the methodology. It's based on this, subject to a bunch of adjustments made by Micah McCurdy.

http://www.silversevensens.com/2014/...tions-with-pip

Ahh, good. I asked earlier what went into his model, but never saw a response. So thank you for this.

So the chart is basically just rolling fenwick, adjusted for schedule. Easy to see why it hates Calgary.

As a proxy for measuring possession, I completely understand why SAT/Corsi and USAT/Fenwick rely on five on five. But if we are going to be predicting outcomes, I think models such as these would be wise to include special teams. One of the reasons why fancystats guys can't understand why the Flames win is because they discount special teams. Over our last 25 games (back to the home loss vs. Florida), the Flames are running at 20% power play (14-70) and 92% penalty kill (44-48). We also take, by far, the fewest penalties in the league, so that limits the opposition's chances.

So while we might be drowning in possession stats in the 80% of the game that is played at even strength, the fact that we are dominating the other 20% right now is a major reason why the Flames defy Corsi/Fenwick-based models. And while I would agree with anyone who says 20%/92% are not figures that are sustainable indefinitely, nor should they be simply discounted as "luck".

[CorsiHockeyLeague]

As I said earlier, that also occurred to me, so I asked Micah about that and he said, first, that he had a separate version of the model that does include some special teams data, but that the predictive benefit of adding it was minor (2+ seasons of sample to add value). Any other tinkering with the model seemed to do more harm than good (I also asked him about factoring in CHIP but he says this just mucked things up), so this is where he landed. It isn't perfect, as should be evident by looking at Minnesota's trend line.

[Textcritic]

Originally Posted by CorsiHockeyLeague

This discussion is pretty similar to the one that came up earlier in the thread when that graph was posted, so no point going over old ground. All I'll say is it's based on a methodology with a record of making correct predictions about future team success. That doesn't mean it's always right and it doesn't mean the Flames won't make the playoffs in spite of the methodology. It's based on this, subject to a bunch of adjustments made by Micah McCurdy.

http://www.silversevensens.com/2014/...tions-with-pip

I'm not certain that I understand everything in the model, even after having read the breakdown. I may have missed it, but does the model merely plug figures for home/away play over the course of the past 25 games, and use these to project figures for each team's remaining home/away games?

If so, then there is a major breakdown in how it accounts for the remainder of the schedule, in which many of the teams in the playoff hunt play games against one another. I have maintained for some time now that there are two factors that should weigh heavily into the Flames favour for making the playoffs, and that is the fact that they hold so many of the tie breakers, but more importantly also that they have far fewer games remaining against teams with whom they are in direct competition for a playoff spot.

In short, and by way of example, LA and Vancouver have three remaining games against each other, and Winnipeg and Vancouver have two. There will be a guaranteed loser in each one of those games, meaning that there are also five opportunities for the Flames to gain position on these teams in the standings. If the Flames were only battling with one team, it wouldn't work out like this, but because there are potentially one of three or four spots that the Flames could land in, then games in which these other teams play one another is actually good for the Flames' chances to make it.

Does the model account for this sort of stuff? If not, it would seem to me to introduce a significant obstacle to its usefulness, and probably explains why the other odds-generators are more closely aligned to one another, and so far from this.

[dissentowner]

I can't believe people are still defending corsi when it has been shown time and again to be garbage. You might as well defend Eklund in trade rumours too. Garbage.

[Street Pharmacist]

Originally Posted by dissentowner

I can't believe people are still defending corsi when it has been shown time and again to be garbage. You might as well defend Eklund in trade rumours too. Garbage.

It hasn't been shown to be garbage. It's very useful and does have predictive value. It just doesn't describe everything.

This kind of absolutism is just as wrong and frustrating as the "corsi explains everything" assertions

[dissentowner]

Originally Posted by Street Pharmacist

It hasn't been shown to be garbage. It's very useful and does have predictive value. It just doesn't describe everything.

This kind of absolutism is just as wrong and frustrating as the "corsi explains everything" assertions

Ok, show me how it has predictive value.

[Igster]

You know what the worst thing about corsi is? Discussions about corsi. Maybe we can keep this to the playoff race in this thread and start a corsi value thread again?

[Street Pharmacist]

Originally Posted by dissentowner

Ok, show me how it has predictive value.

Take the half way point of the season, say January 8th

Here's your top 10 Corsi 5v5 teams:
LA
TB
CHI
DET
NYI
NSH
MIN
BOS
WPG
FLA

By using this metric to simply determine who's going to make the playoffs, we'd be at least 80% correct with a good chance of 90%

[EldrickOnIce]

Originally Posted by dissentowner

Ok, show me how it has predictive value.

It is both statistically and intuitively valuable.
There is a ton of info/analysis that shows it has predictive value.
The disconnect is that one side will tell you it is everything, the other tells you it means nothing. Reality is somewhere in the middle.
Intuitively, you are less likely to win (in any term, long or short) if you are consistently outshot and outplayed. That is completely obvious.

Problem is guys like Lambert, CHL, etc. will tell you it is the only thing that matters, to the point where wins and losses are wrong and don't reflect which team's are better. Which is complete nonsense.

[CorsiHockeyLeague]

Originally Posted by EldrickOnIce

Problem is guys like Lambert, CHL, etc. will tell you it is the only thing that matters, to the point where wins and losses are wrong and don't reflect which team's are better. Which is complete nonsense.

I'm trying to figure out where I said this. It's surprising to hear that I did, because I don't remember saying that, and I'm pretty sure I've in fact said the exact opposite on more than one recent occasion.

[EldrickOnIce]

Originally Posted by Street Pharmacist

Take the half way point of the season, say January 8th

Here's your top 10 Corsi 5v5 teams:
LA
TB
CHI
DET
NYI
NSH
MIN
BOS
WPG
FLA

By using this metric to simply determine who's going to make the playoffs, we'd be at least 80% correct with a good chance of 90%

You can also look at the bottom of the standings. These are the lowest corsi teams on the season:

Buffalo, Colorado, Toronto, Columbus, New Jersey, Edmonton and Arizona.... but mixed in there are Calgary and Montreal.

So predictive, absolutely? But far from absolute

[EldrickOnIce]

Originally Posted by CorsiHockeyLeague

I'm trying to figure out where I said this. It's surprising to hear that I did, because I don't remember saying that, and I'm pretty sure I've in fact said the exact opposite on more than one recent occasion.

Then I misread one of your posts, and I am not correct in lumping you in there.

This is what I was referring to

Originally Posted by CorsiHockeyLeague
Please explain why you'd take that view. The statement "You're wrong, clearly X team is better than Y team because they're higher in the standings" barely deserves a response even if you don't believe in any "advanced" stat. It's clearly wrong, and we've just been talking about how obvious that is.

[mrkajz44]

Originally Posted by CorsiHockeyLeague

As I said earlier, that also occurred to me, so I asked Micah about that and he said, first, that he had a separate version of the model that does include some special teams data, but that the predictive benefit of adding it was minor (2+ seasons of sample to add value). Any other tinkering with the model seemed to do more harm than good (I also asked him about factoring in CHIP but he says this just mucked things up), so this is where he landed. It isn't perfect, as should be evident by looking at Minnesota's trend line.

I guess this is part of the issue with the model though (and any model for that matter). For most teams, maybe adding special teams doesn't make a big difference. But the Flames have allowed something like 2 or 3 PPG against in their last 15 or so games (sorry, don't have the time to look up the actual stats). So, while in general adding the special teams doesn't seem to add value for the model overall, it could greatly affect a team or two, such as the Flames.

[getbak]

Originally Posted by Street Pharmacist

Take the half way point of the season, say January 8th

Here's your top 10 Corsi 5v5 teams:
LA
TB
CHI
DET
NYI
NSH
MIN
BOS
WPG
FLA

By using this metric to simply determine who's going to make the playoffs, we'd be at least 80% correct with a good chance of 90%

Of the 16 teams who were in a playoff position on January 8, 14 are still in a playoff position (Calgary and Minnesota are the only two who were on the outside then who are in now). It's likely that 14 or 15 of the teams who were in a playoff position on January 8 will make the playoffs (San Jose is likely to drop out for Minnesota, and Calgary may bounce out LA, Winnipeg, or Vancouver).

That's 87.5 - 93.75% accuracy, so why bother trying to find some other solution, when the most basic formula already exists and is at least as accurate?

[Street Pharmacist]

Originally Posted by getbak

Of the 16 teams who were in a playoff position on January 8, 14 are still in a playoff position (Calgary and Minnesota are the only two who were on the outside then who are in now). It's likely that 14 or 15 of the teams who were in a playoff position on January 8 will make the playoffs (San Jose is likely to drop out for Minnesota, and Calgary may bounce out LA, Winnipeg, or Vancouver).

That's 87.5 - 93.75% accuracy, so why bother trying to find some other solution, when the most basic formula already exists and is at least as accurate?

I agree with you, I'm only giving a simple proof to dissentowner that corsi does have predictive power

[dissentowner]

Originally Posted by Street Pharmacist

I agree with you, I'm only giving a simple proof to dissentowner that corsi does have predictive power

And yet Montreal and Calgary are poor Corsi teams but one is the best team in the East and the other is in the thick of the playoff race. It has predictive power but so does a magic 8 ball.

[troutman]

Originally Posted by Street Pharmacist

By using this metric to simply determine who's going to make the playoffs, we'd be at least 80% correct with a good chance of 90%

Is is any more predictive than other indicators?

Ex. Goals against, goal differential, road record, PP% - PK%, standings

I recall road record being a fairly reliable predictor of playoff success.