Article: How the Calgary Flames are Laughing at Advanced Stats

[EldrickOnIce]

Originally Posted by Enoch Root

As for Corsi predicting approximately 70% of the playoff teams... in sports, in any given year, predicting 70% of the playoff teams is an absolute slam dunk, and should be considered the very basest of baselines.

Show me a model or metric that can predict to harder 30%, and then I will be impressed.

I bet i can predict 11/16 (70%) of the playoff teams each year. On average, at least 70% of teams that made the playoffs last year will again the next year. Playoff teams will be reflected by goal differential around 90% of the time.
Burke still said it best, 'supportive but not illuminating'. Every team looks for an edge. Some of that edge will come from statistical analysis, but any advantage sure won't come from score adjusted corsi close. The Oilers grabbed the free agent corsi darlings last summer, and we know how that has turned out.

[Resolute 14]

The "predictive power" of any stat is a very bad argument, in my view because of a lack in context and because stats simply don't predict anything by themselves. e.g.: Last year's stats are frequently not predictive for this year or next because of high roster turnover. Even a team like Pittsburgh where their core is still young, stable, and producing, can go from 51 wins 13 points clear of second in the division to the risk of being a wild card team in just one year.

So how far into a season do we need to go before statistics can capably "predict" something? The Hockey News has consistently demonstrated that the dartboard approach is good for 70%. If LA knocks either us or Winnipeg out, they will have hit on 11/16 (69%) in their pre-season guesses. If LA fumbles somehow, it will fall to 63%. They went 11/16 last year too.

Shot attempts/Corsi/Fenwick don't predict anything. All they do is add another point of data to the guessing game, and that is largely why their "predictive power" is - at best - marginally better than spit balling. Because it is still the subjective interpretation of the person using the stats that defines the prediction. And I don't care what statistics or methods you use, the only people who predicted the Flames would be in this battle before the season started did so because they sought attention, not because they actually believed it.

[Enoch Root]

And yet people build predictive charts based on Corsi-close and other people defend their predictability vehemently.

And around and around we go.

[CorsiHockeyLeague]

If you want to make predictions, you can either base them on something, or nothing. If you're basing them on something quantifiable, the best available option is 5v5 fenwick (or fenclose if you have a large enough sample).

[_Q_]

Originally Posted by Enoch Root

And yet people build predictive charts based on Corsi-close and other people defend their predictability vehemently.

And around and around we go.

Yeah and I think that's the core of the pro-corsi vs anti-corsi crowd. Corsi are just statistics. You have to be fairly ignorant to hate on statistics. But you also have to be equally as ignorant to make statements like "team X will miss the playoffs because the Corsi says so".

[northcrunk]

The only thing you can accuratly predict is you cannot accuratly predict anything. Unless you have a time machine or are Dr.Who.

[CorsiHockeyLeague]

Originally Posted by northcrunk

The only thing you can accuratly predict is you cannot accuratly predict anything. Unless you have a time machine or are Dr.Who.

I predict that the Montreal Canadiens will make the playoffs.

And it's just "the Doctor", please.

[V]

The anti-stats crowd is hilariously uninformed, but the pro-stats side is also overly zealous. You can't use a proxy and overinflate its importance. There will come a time soon when the proxy won't be needed and then maybe we'll have useful data. Right now it's just a shade.

But the idea that you can't use stats because the data is inaccurate is also comical. We aren't working out of a textbook here, this is the real world. Data is dirty. If we were worried about working with pristine data there wouldn't be such a thing as statistics. It's also comical how the anti-stats crowd tries to discredit the stats crowd by rewriting their arguments into something ridiculous, because that's always worked in a debate. Something like

team X will miss the playoffs because the Corsi says so

is so painfully ignorant that I'm always amazed that the stats guys even have any will to 'discuss' this topic with them. I mean, what the hell is the point? Let the Neanderthals be.

[CorsiHockeyLeague]

Originally Posted by V

The anti-stats crowd is hilariously uninformed, but the pro-stats side is also overly zealous. You can't use a proxy and overinflate its importance. There will come a time soon when the proxy won't be needed and then maybe we'll have useful data.

100% agree, but I don't think people are overinflating its importance for the most part (maybe sensationalist guys like Lambert). It's really more of an accusation you hear from the anti-stats guys: "You think this explains everything! Why even play the games? Why not just hand the Cup to the team with the highest CORSI snicker snicker?"

Last sentence there is dead on though. Personally I can't wait for SportVU, we'll get to start all over again in determining what helps teams win and what doesn't, only on a much more granular scale.

But the idea that you can't use stats because the data is inaccurate is also comical. We aren't working out of a textbook here, this is the real world. Data is dirty. If we were worried about working with pristine data there wouldn't be such a thing as statistics. It's also comical how the anti-stats crowd tries to discredit the stats crowd by rewriting their arguments into something ridiculous, because that's always worked in a debate on the internet.

fixed.

[Resolute 14]

Originally Posted by _Q_

Yeah and I think that's the core of the pro-corsi vs anti-corsi crowd. Corsi are just statistics. You have to be fairly ignorant to hate on statistics. But you also have to be equally as ignorant to make statements like "team X will miss the playoffs because the Corsi says so".

It's not really ignorance though, but rather overconfidence. A prediction is, by definition, a guess.

"Based on Corsi, I think the Flames will miss the playoffs."
"Based on goal differential, I think the Flames will make the playoffs."

Neither prediction is rooted in ignorance. Both are supported by quantifiable statistics that carry historical relevance. But one of them is going to be wrong.

[SofaProfessor]

I think, in time, we will find out how different teams are using these advanced stats. I wouldn't be surprised if some teams have their own "modified corsi" for lack of a better term. We keep talking about quality of shots versus shot attempts and how, for example, a shot coming off a 2 on 1 rush is a better quality shot than a shot coming from along the boards in a 5 on 5 situation. I'm sure there are teams out there with an advanced stats guy on staff that has come up with some sort of formula that takes into account quality of shots. I dunno. I'm not smart enough to explain myself in great detail. But I think it would be a mistake to assume all NHL teams are using these advanced stats in the same way or even calculating them the same way.

[_Q_]

See its this attitude by the Corsi crowd that annoys the rest of us. Neanderthals? Are you f'ing kidding me?

Originally Posted by Resolute 14

It's not really ignorance though, but rather overconfidence. A prediction is, by definition, a guess.

"Based on Corsi, I think the Flames will miss the playoffs."
"Based on goal differential, I think the Flames will make the playoffs."

Neither prediction is rooted in ignorance. Both are supported by quantifiable statistics that carry historical relevance. But one of them is going to be wrong.

Sure, I can get behind this, but it's funny you bring up goal differential since it's been more effective at predicting which teams are in playoff position this season than any advanced stat (15/16 as opposed to 11/16).

[heep223]

Sigh. I should have known this would happen.

[Resolute 14]

Originally Posted by heep223

Sigh. I should have known this would happen.

Heh, yeah. It's more interesting than 100 replies of "yup", though.

[V]

Well, duh. What did you expect??

See its this attitude by the Corsi crowd that annoys the rest of us. Neanderthals? Are you f'ing kidding me?

I'm far from a Corsi guy. Not sure what you found in my posting history that would point you in that direction.

[CorsiHockeyLeague]

Originally Posted by _Q_

Sure, I can get behind this, but it's funny you bring up goal differential since it's been more effective at predicting which teams are in playoff position this season than any advanced stat (15/16 as opposed to 11/16).

This is utter nonsense as an argument, though. Points come from wins, which come from outscoring your opponent. It's essentially tautological to say "goal differential determines playoff seeding"; you might as well go one step further and say 16/16 playoff teams are predicted by their win/loss record. It doesn't "predict" playoff seeding, it describes playoff seeding, because goals for percentage cannot be relied upon to stay the same over time.

Put another way, if asked, "if I want my team to make the playoffs, what should they be good at", the person who understands analytics will say, "it's no guarantee, but the best thing would be for them to be good at possessing the puck". Your response appears to be, "make them good at scoring more goals than the other guys". That's completely unhelpful.

[Resolute 14]

Originally Posted by CorsiHockeyLeague

Put another way, if asked, "if I want my team to make the playoffs, what should they be good at", the person who understands analytics will say, "it's no guarantee, but the best thing would be for them to be good at possessing the puck". Your response appears to be, "make them good at scoring more goals than the other guys". That's completely unhelpful.

Which explains why mudcrutch became the buttmonkey of the site by penning a treatise called "the value of outscoring: a primer". It was completely unhelpful.

That said, I agree with your premise. "Score more goals than the other guys" is the outcome, not the process. Being a strong possession team is one method that could lead to that outcome. So too is having the world's best goaltender.

[edslunch]

Nice article, but I hate the way he starts, like all advanced stats article, by saying the flames are a bad team. They're not bad - in some areas they are quite good, in others middle of the road, and sure in possession stats quite poor.

[_Q_]

Originally Posted by V

Well, duh. What did you expect??



I'm far from a Corsi guy. Not sure what you found in my posting history that would point you in that direction.

Well calling people that are arguing the validity of advanced stats Neanderthals kind of points me in that direction don't you think?

Anyways, that's neither here nor there. Bring something valid into this debate rather than calling people names, because that sure as hell doesn't work.

[CorsiHockeyLeague]

Originally Posted by Resolute 14

Which explains why mudcrutch became the buttmonkey of the site by penning a treatise called "the value of outscoring: a primer". It was completely unhelpful.

Yeah, looking back at stuff written a decade ago, you're going to find some "wtf was that guy thinking" articles. Dellow wrote a metric ton of stuff, it seemed to be just whatever came into his mind as "this might be interesting to look at", some of which ended up being very enlightening and some not, but he wrote it all out anyway having done the work. I do wish MC79Hockey still existed as a site because there was a treasure trove of good ideas there.