Article: How the Calgary Flames are Laughing at Advanced Stats

[heep223]

Still, it'd be an improvement on the current stats I think. As it currently stands, a Bollig snapshot from the blue line with no traffic is the same "possession-wise" as a quality Monahan scoring chance from the slot. Makes no sense at all.

[CorsiHockeyLeague]

Originally Posted by heep223

Why can't draw a box from the posts, to the dots, to the high slot and shots/attempts from that box are worth more? Where shots have been attempted from is already tracked with those heat maps, I've never understood why the stats guys couldn't use that to measure shot quality and incorporate it into a stat.

This has been done literally for more than half a decade. It's called home plate.

http://flamesnation.ca/2015/1/7/revi...half-way-point

Here is why it doesn't really improve current stats: http://nhlnumbers.com/2012/6/26/shot...nd-shot-totals

That latter article is about shot quality which was one of the most discussed topics for several years. There are dozens of articles about it. It's not like no one noticed that a shot attempt from the point counts the same as one from the slot for corsi purposes.

[Enoch Root]

No one is saying that no one has noticed, or that research hasn't been done. Some research has been done, but no one has yet been able to push things materially forward.

That DOES NOT mean that the current, insufficient stats should just be accepted because 'there's nothing better'.

It also does not mean that the questions shouldn't continue to be asked.

They continue to be asked because, so far, what we've got isn't good enough.

Continuing to ask the questions doesn't mean people are uninformed, it is simply part of the scientific process.

[CorsiHockeyLeague]

They shouldn't continue to be asked because they've been answered!

If you read that article and think, "once again, I persist in requesting that shot quality data be added to the existing statistics because it will make a difference or improve them in some way", you didn't understand the article. Or the many, many others that have reached the same conclusion.

[PeteMoss]

I will say this... I do agree with Tango in that types of shots should be weighted in a certain way.

Slugging percentage is a better stat than batting average in baseball because of weighting a HR more than a single.

I doubt it tremendously changes anything but a one-timer from the slot should be weighted more heavily than a dump in from centre.

[CorsiHockeyLeague]

I honestly can't tell if I'm being trolled at this point.

I'm out.

[PeteMoss]

And stuff like this - https://jenlc13.wordpress.com/2015/0...ve-breakdowns/ (who is being targeted while teams break-in to the offensive zone) is actually advanced stats work and much more helpful for evaluating players than shot differentials

[PeteMoss]

Originally Posted by CorsiHockeyLeague

I honestly can't tell if I'm being trolled at this point.

I'm out.

Feel free to read this when you get over yourself. http://tangotiger.com/index.php/site...tial-aka-tango

[Textcritic]

So, back to the article, which I thought was excellent.

What stood out to me most was this: McIndoe flags four teams from the past few seasons as prior examples of those that defied the odds for a time, but ultimately conformed to the metrics. the Dallas Stars, the Minnesota Wild, the Colorado Avalanche, and the Toronto Maple Leafs. I have said before that I will not be a whit surprised to see the Flames regress and repeat this same pattern next year. But, there is still no other way to look at what has happened this season as good news.

What do 3/4 of those other teams have in common? They are either presently very good teams, or they look like they will be very soon. The Wild look like they could win a round or two this year. The Avalanche and the Stars both look like playoff teams that had a bad couple of months. The Maple Leafs are terrible.

What does this mean for the Calgary Flames? If I had to guess, I would say that being more like the Wild, Avalanche and Stars in their current franchise trajectory, it suggests that while they may stumble a couple more times in the next year or two, there are reasons to expect that they will find success sooner than later.

Maybe that is something that can be said for bad possession teams who do find some unexpected success? That they are doing something right now which will bode well for them in the future...Well, for all of them except for the Leafs.

[heep223]

Originally Posted by CorsiHockeyLeague

I honestly can't tell if I'm being trolled at this point.

I'm out.

Stop being so dramatic. If you're out, then stop posting. Why not just adopt the attitude that you're helping people understand the statistics and that we're all on the same side.

As per the article you posted, thank you, it was educational. However scoring chances and shot attempts are correlated at .8. That's telling, but not perfect. It further dilutes the stat's ultimate predictive power. If we had a stat that actually used a weighting system for shot quality that would improve it.

[Coach]

Originally Posted by CorsiHockeyLeague

They shouldn't continue to be asked because they've been answered!

If you read that article and think, "once again, I persist in requesting that shot quality data be added to the existing statistics because it will make a difference or improve them in some way", you didn't understand the article. Or the many, many others that have reached the same conclusion.

There's still questions and scenarios on top of what is looked at in the article. The article is assuming that a shot which is a scoring chance is "quality shot", which is not necessarily true. What is classified as a shot vs a scoring chance is hugely subjective and again does not take into account the quality of the individual player. Some players shoot harder than others. Some players are more accurate than others. Who are they playing against? Is a scoring chance more important or better if it's done while Weber/Josi are defending vs Schultz/Ference?

The stats assume this stuff is equal and it's just not. Every player is different and affects the game in a completely different way. We're not saying that the possession stats aren't meaningful or useful, just that they don't tell the whole story (which you admit they don't). So some are on the hunt for the full story, and are of the opinion that we may never be able to isolate the one or two stats that CONTRIBUTE to winning. Attempting to win by improving these stats will not work, as they don't contribute to winning, they just tell us that the attributes of this team (or player) leads them to posses the puck more. What those attributes are that lead to puck possession are largely variable. Even something like take aways doesn't tell the whole story. Are those take aways in the ozne on the forcheck? Dzone on the backcheck? One player may have a bunch of for checking take aways but none on the back check.

The point is that hockey is likely the most variable sport, making it difficult to track and isolate the affect of individual players. Largely because we (and the sport overall) is still learning what those attributes are and they change with every change of rule or equipment. A couple years ago we'd be talking about how you need an elite 50 goal scorer to be a top team. Now we say you need a collection of two way players, giving up individual offence and scoring by committee. It changes all the time.

And stuff like this:

That's not much of a trend. With an R^2 of 0.049, the correlation is strong enough to be statistically significant, but weak enough to be insignificant in practice. If a guy is at 0.9 this year (one of the worst in the league), you can't even say that he'll be below average next year -- just that he won't be one of the very best.

is annoying to those of us who don't believe it's all figured out. He "WONT" be one of the best? Well he could. This statistcal analysis has no impact or bearing on what may happen next year. Sure you can use it to try and predict, and you may be right in most cases, but to say WONT, is wrong. Just like saying the Flames WILL regress. They may regress, and it's statistically likely that they do. But that doesn't mean they will.

[Street Pharmacist]

Originally Posted by CorsiHockeyLeague

They shouldn't continue to be asked because they've been answered!

If you read that article and think, "once again, I persist in requesting that shot quality data be added to the existing statistics because it will make a difference or improve them in some way", you didn't understand the article. Or the many, many others that have reached the same conclusion.

I've read that article a few times now, but it makes the same error that all "sample size" conclusions do: on the whole is correct, but it leads to erroneous conclusions about individuals.

I'll use a healthcare example. If we look at the risk of rhabdomyolysis in those taking statins like Lipitor and Crestor, we'd see it's a dose related side effect. We then can say your curable If we didn't look forget into subgroups (specifically genetic or ancestral) we wouldn't have noticed those of Asian descent (mostly Japanese) do not metabolize the drug as fast and thereby require lower doses and have higher risk of rhabdomyolysis at lower doses.

More shots = more scoring chances generally, yes. That doesn't mean outliers by default are random. Perhaps there's more at play? The only time shooting % is found generally unsustainable is using whole season data and comparing season to season. I'd say that's apples and oranges, because roster turnover means you simply cannot compare and trust the results.

[Enoch Root]

Originally Posted by CorsiHockeyLeague

They shouldn't continue to be asked because they've been answered!

If you read that article and think, "once again, I persist in requesting that shot quality data be added to the existing statistics because it will make a difference or improve them in some way", you didn't understand the article. Or the many, many others that have reached the same conclusion.

The questions have been addressed. But they have NOT been answered. Until a model exists that definitively captures the data and is 100% accurate in describing and differentiating, nothing has been answered.

Many articles have been written, but no one has been able to push things significantly forward yet. So they have not been answered, unless one believes that all useful information is in fact now being captured and there is no more information to garner.

Attempting to differentiate shot quality, but discovering that it doesn't help, is not definitive proof that shot quality doesn't matter. It only means that , so far, the information and techniques/methods aren't good enough to push the science forward and qualify as a bona fide improvement. That does not mean that there aren't improvements still to be made.

[PeteMoss]

Originally Posted by Enoch Root

The questions have been addressed. But they have NOT been answered. Until a model exists that definitively captures the data and is 100% accurate in describing and differentiating, nothing has been answered.

Many articles have been written, but no one has been able to push things significantly forward yet. So they have not been answered, unless one believes that all useful information is in fact now being captured and there is no more information to garner.

Attempting to differentiate shot quality, but discovering that it doesn't help, is not definitive proof that shot quality doesn't matter. It only means that , so far, the information and techniques/methods aren't good enough to push the science forward and qualify as a bona fide improvement. That does not mean that there aren't improvements still to be made.

According to these war-on-ice guys, scoring chances are (slightly) better than corsi for determining future goals: http://blog.war-on-ice.com/scoring-c...er-than-corsi/

[heep223]

Originally Posted by PeteMoss

According to these war-on-ice guys, scoring chances are (slightly) better than corsi for determining future goals: http://blog.war-on-ice.com/scoring-c...er-than-corsi/

That is very interesting. I like how they've defined scoring chance and to me this is actually a lot closer to an advanced stat than shot differentials. I might have missed it, but are they developing a set of statistics around this? I'd be very interested to see them. It might go a long way to explain the Flames continuing defiance of the possession stats - I would assume their scoring chance to shot attempt differential is very high. ie. they tend to shoot when they have a good chance to score, moreso than other teams around the league.

A rules-based scoring chance differential would be really cool to see also, the current way of looking at chances is way too subjective.

[Street Pharmacist]

Originally Posted by heep223

That is very interesting. I like how they've defined scoring chance and to me this is actually a lot closer to an advanced stat than shot differentials. I might have missed it, but are they developing a set of statistics around this? I'd be very interested to see them. It might go a long way to explain the Flames continuing defiance of the possession stats - I would assume their scoring chance to shot attempt differential is very high. ie. they tend to shoot when they have a good chance to score, moreso than other teams around the league.

A rules-based scoring chance differential would be really cool to see also, the current way of looking at chances is way too subjective.

Actually no. Calgary's corsi% is about the same as the scoring chance%. In the end, Calgary just finishes a higher percentage of their chances. By skill or chance is the argument, but they generally get quite a few less chances than the opposition

[PeteMoss]

Originally Posted by heep223

That is very interesting. I like how they've defined scoring chance and to me this is actually a lot closer to an advanced stat than shot differentials. I might have missed it, but are they developing a set of statistics around this? I'd be very interested to see them. It might go a long way to explain the Flames continuing defiance of the possession stats - I would assume their scoring chance to shot attempt differential is very high. ie. they tend to shoot when they have a good chance to score, moreso than other teams around the league.

A rules-based scoring chance differential would be really cool to see also, the current way of looking at chances is way too subjective.

It was already shown in another thread that if you don't limit Corsi/Fenwick to 5 on 5 like most do, the Flames while still below average move from crazy outlier to 'a team getting some breaks' (particularly if you use Fenwick).

http://war-on-ice.com/teamtable.html...0&tablegroup=4

[heep223]

Originally Posted by PeteMoss

It was already shown in another thread that if you don't limit Corsi/Fenwick to 5 on 5 like most do, the Flames while still below average move from crazy outlier to 'a team getting some breaks' (particularly if you use Fenwick).

http://war-on-ice.com/teamtable.html...0&tablegroup=4

Is that Buffalo that's so far in the bottom left corner that the text is unreadable? lol

[PeteMoss]

Originally Posted by heep223

Is that Buffalo that's so far in the bottom left corner that the text is unreadable? lol

Yes I would assume so...

Click to 'Tabular View' to see the table... not sure why it defaults to a graph.