Article: How the Calgary Flames are Laughing at Advanced Stats

[Enoch Root]

Oh, and great article

[JiriHrdina]

Originally Posted by CorsiHockeyLeague

Yeah but it's a pain in the ass for me or anyone else to have to explain the same things and post the same links to the same resources over and over and over.

Not to mention a lot of the people asking the questions aren't legitimately interested in answers, they just want to raise perceived (almost always nonsensical and ignorant) objections because they reject any statistic or analytical method that doesn't say good things about the team they cheer for.

If nothing else, just recognize that whatever brilliant hole you've found that totally discredits these stats, no, you are not the first person to think of it, and yes, it has been discussed and analyzed at length.

Continue to talk down to people. That'll help further advanced stats for sure

[CorsiHockeyLeague]

I've been about as accommodating and helpful as anyone could be expected to be in light of the attitude from many (by no means all) posters around here. Maybe the issue is more the double standard that exists in that respect.

[Bend it like Bourgeois]

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.

Well put.

There are some really interesting insights from stats. There is nothing more predictive about them than any other insight though.

Home ice advantage is well established.

Yet a good team with a strong road record is just a good team. Home ice be damned.
A bad team with a strong road record though is a story. Why are they road warriors? Is their home or road record the real them? Good conversations.

Nobody goes full weenie and suggests a good road record is unsustainable, based on unassailable math. They take it for what it is, an insight worth a conversation.

[CorsiHockeyLeague]

Originally Posted by Enoch Root

Many of the posters here who question these stats have already read those articles, understand the issues, and yet - implausibly - still question and challenge the conclusions and claims.

Why? Because that is the proper way to go about something that is so far from being complete that it is still closer to complete ineffectuality.

Let's just take it as read that you're right and the people asking these questions - like the shot quality one earlier in this thread - have in fact read the myriad articles addressing their questions and explaining why they should not, or to what extent they should, create an error bar on the usefulness of shot attempts as a statistic.

That's clearly not the case, based on the tone of the questions ("how can anyone think Corsi is useful when it doesn't take into account shot quality? It treats a shot attempt from the point the same as one from the slot!") but let's just assume it is.

Raising issues, having those issues fully addressed, and then raising them again isn't a way to go about anything. How could it be?

[JiriHrdina]

Originally Posted by CorsiHockeyLeague

I've been about as accommodating and helpful as anyone could be expected to be in light of the attitude from many (by no means all) posters around here. Maybe the issue is more the double standard that exists in that respect.

You dismiss people who don't support your stance on advanced stats and seem to quickly lose patience who ask legitimate question.

If you want people to buy-in to anything that is somewhat new, at least to the broader fan base, then patience will be required.

Arrogance and being dismissive won't help.

[CorsiHockeyLeague]

I think you'll find if you fairly assess the people who are anti-stats on here display a considerably more arrogant and dismissive attitude - also generally mocking. I have answered a number of legitimate and honest questions. As has Street Pharmacist and a couple of others.

However, many others are simply "here's what I think is stupid about corsi". It's not an instance of someone wanting more information or wanting to know why the issue they've raised is right or wrong - it's precisely dismissive for the sake of being dismissive. It's "the Flames are bad at this so I would like it discredited".

It's also exasperating, so I can either get into a pointless discussion with someone who doesn't want a discussion or I can be dismissive and say "multiple people have written about why this criticism is wrong, so try looking for them."

[JiriHrdina]

It is a vicious cycle but if you want the movement to gain more acceptance the advanced stats community has to figure out how to have the conversation without the attitude.
It was the same problem that our old buddy muddy had - though far worse with him.

I really enjoy the analytics guy that 960 has on in the afternoons. He has interesting things to say, doesn't trot out things like "the Flames will regress next season" and very patiently explains what it all means. Kerr asks him direct questions and he responds perfectly - addressing the objections. He is a guy that clearly believes in the advanced stats but uses them to supplement how he watches the game.

[Lanny_McDonald]

@CorsiHockeyLeague

Garbage in is garbage out. The league acknowledges inconsistency in data collection which means data sets are flawed. How do you think that affects reliability in your statistical analysis? I really think that many of the people who embrace advanced stats don't have a high degree of understanding of the subject matter. There is a need for some advanced stats classes, which would change your understanding of the topic and make you realize just how flawed the assumptions that are made really are. Hockey is a very complex game that relies on chaos, which is next to impossible to model. The more people try the more failure they achieve.

[starseed]

I watch the game for the random streaks and events. If the standings were immune to them, I would care much less.

I am still not convinced that Corsi has figured out how to determine a team's effectiveness. How about goal differential? It is 94% certain to predict a team's place in the standings! That is better than 69%!

When you have even a handful of teams in the league whose playbook leads to routinely bucking the Corsi predictions, I think you better go back to the drawing board.

[Slava]

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.

This is exactly what I wanted to say when I read that 70% figure. I think that most people on this board could get 70% sitting and looking things over pretty casually in October before the puck drops. That's not a badge of honour!

[Ryan Coke]

CHL I actually tend to agree that many people don't want to learn, and are mostly cheering against advanced stats because some people use them to predict doom for the team they cheer for.

I have absolutely nothing against the stats, and I enjoy learning more about some of their subtleties and statistical correlations. However often the stats supporters turn into cheerleading for their cause, which is the stats they believe in.

It would be refreshing to hear your views on the weaknesses of the stats, because of course there are many. And the predictive powers of these stats are also far from certain. And yet too often the stats supporters refuse to acknowledge that.

Can you imagine if there was someone who called themselves "+\- hockey league" and then with a fans passion told everyone why +\- was such an awesome stat?

It's stats....they are what they are, it's the conclusions people draw from them that is the issue.

[bob-loblaw]

Nm.

[CorsiHockeyLeague]

Originally Posted by JiriHrdina

It is a vicious cycle but if you want the movement to gain more acceptance the advanced stats community has to figure out how to have the conversation without the attitude.

Honestly I think the "corsi war" is over. That should have been obvious when the Leafs bought extraskater (along with the half-dozen other guys whose sites went dark). A large number of the best analytics writers now work for NHL teams or TSN/Sportsnet. It's basically mainstream and writing these metrics off is basically non-viable. They're on NHL.com, ffs. Badly implemented, but there nonetheless. The "gaining more acceptance" stage is pretty much over even if some people are still getting up to speed on how to use these things.

I really enjoy the analytics guy that 960 has on in the afternoons. He has interesting things to say, doesn't trot out things like "the Flames will regress next season" and very patiently explains what it all means. Kerr asks him direct questions and he responds perfectly - addressing the objections. He is a guy that clearly believes in the advanced stats but uses them to supplement how he watches the game.

I don't listen to radio so I don't know who that is - do you know his name? There are some heavyweights in Calgary and Alberta in analytics. There have been multiple analytics conferences; one in the dome later last year. Justin Azevedo and Ryan Pike, Kent Wilson especially is a heavyweight. Hell, Rob Vollman lives here. If it's any of those guys, for sure, they know their stuff.

I would say if you like listening to this stuff on the radio rather than reading about it, listen to Travis Yost's show on TSN on the weekend, or just pull the episodes up on TSN.ca.

[CorsiHockeyLeague]

Originally Posted by Ryan Coke

It would be refreshing to hear your views on the weaknesses of the stats, because of course there are many. And the predictive powers of these stats are also far from certain. And yet too often the stats supporters refuse to acknowledge that.

First, as stats, they don't have "weaknesses"; rather there are things that they describe and things they don't. Anyone who thinks that the point is "all you need to know in player evaluation is corsi" doesn't get it.

Further, anyone who says the predictive power of the stats is perfect doesn't understand them. This is a game played by human beings on an ice surface with a bouncy rubber disc. There's absolutely no way to perfectly predict anything. However, among stats that correlate to winning, they have the best combination of correlation and reliability - i.e., your ability to generate shot attempts at a higher rate than the opposition contributes significantly to winning, and if you've out-attempted your opposition in the past, you're likely to continue to do so in the future. Other stats fall short in one area or the other - either they don't actually contribute to winning hockey games (i.e. outhitting your opponent or blocking more shots), or they aren't repeatable going forward (i.e. even strength save percentage).

[HOWITZER]

Originally Posted by CorsiHockeyLeague

First, as stats, they don't have "weaknesses"; rather there are things that they describe and things they don't. Anyone who thinks that the point is "all you need to know in player evaluation is corsi" doesn't get it.

Further, anyone who says the predictive power of the stats is perfect doesn't understand them. This is a game played by human beings on an ice surface with a bouncy rubber disc. There's absolutely no way to perfectly predict anything. However, among stats that correlate to winning, they have the best combination of correlation and reliability - i.e., your ability to generate shot attempts at a higher rate than the opposition contributes significantly to winning, and if you've out-attempted your opposition in the past, you're likely to continue to do so in the future. Other stats fall short in one area or the other - either they don't actually contribute to winning hockey games (i.e. outhitting your opponent or blocking more shots), or they aren't repeatable going forward (i.e. even strength save percentage).

Sounds like you just disproved yourself

All jokes aside, statistical analysis does have flaws, in the same way that scientific theories that we hold as "true" remain theories instead of law. Scientific theories (much like advanced statistics) can still be disproven and research into doing so is often encouraged as progress. What I think people are getting at is that they want to hear you offer an insider's perspective on where you think Corsi, Fenwick, etc. have shortcomings, much in the same way you have critiqued more traditional measures. Showing inward critique of the numbers/methods you use will give people confidence that you are aware of the information that you are presenting and can confidently defend the merits of it whilst still acknowledging the shortcomings of the data.

[CorsiHockeyLeague]

You have to be more specific as to what you mean by shortcomings - corsi describes shot attempts at the opposing net. If you're saying that it has a "shortcoming" because, for example, it is not a crystal ball that will tell me who will win the Flames' next 5 games, okay, sure. But in terms of describing what it intends to describe - i.e. whether a team had spent more time in the opponent's end with the puck than their opponent spent in theirs over a series of games - it's really quite good at that.

[Ryan Coke]

Exactly, the stat is the stat. But obviously to anyone reading these debates, there are often predictions that are based on the stat. That is the issue. Have you honestly missed that?

So either you don't feel the stat has any predictive usefulness, or you should be able to discuss the limitations of any predictions based on it.

[FlameZilla]

Originally Posted by heep223

Please, please don't turn this thread into another advanced stats debate. I really just wanted to share this Grantland article because it's one of the best I've read about this year's Flames.

http://grantland.com/the-triangle/ho...dvanced-stats/

[CorsiHockeyLeague]

Originally Posted by Ryan Coke

Exactly, the stat is the stat. But obviously to anyone reading these debates, there are often predictions that are based on the stat. That is the issue. Have you honestly missed that? So either you don't feel the stat has any predictive usefulness, or you should be able to discuss the limitations of any predictions based on it.

Obviously I think it has some predictive usefulness:

Originally Posted by Me, two posts ago, responding to you

However, among stats that correlate to winning, they have the best combination of correlation and reliability - i.e., your ability to generate shot attempts at a higher rate than the opposition contributes significantly to winning, and if you've out-attempted your opposition in the past, you're likely to continue to do so in the future. Other stats fall short in one area or the other - either they don't actually contribute to winning hockey games (i.e. outhitting your opponent or blocking more shots), or they aren't repeatable going forward (i.e. even strength save percentage).

As for its limitations, it does not have a perfect correlation to win% nor a 100% repeatability. No stat does.