How Screwed are the Flames?

[troutman]

Originally Posted by CorsiHockeyLeague

Yes they do. It's called score effects. Any decent stats site (stats.hockeyanalysis.com) will provide you with data for different game states - 5 on 5, 5 on 4, 4 on 5, up 1, down 1, tied, close, etc. This is actually fundamental to using these stats, and a point of contention - at what stage does the sample size get big enough for score close situations that the signal outpaces the noise.

Thanks, I learned something.

[heep223]

Originally Posted by CorsiHockeyLeague

I could again end up being wrong (see again my signature), but that doesn't make the bet unsound.

Is there an advanced stat for "team of destiny" (as per your sig)? I'm starting to think more and more that this year's iteration of the Flames is it...

Maybe the advanced stat world needs to create "character %". Something to do with come from behind wins, clutch goals, strength of leadership, closeness of the team....some way to quantify a team with a ton of heart. I started off that thought tongue-in-cheek, but actually I'm not being facetious. IMO these things (along with coaching) are just as important as anything else that we've been discussing. Hurry, someone figure that out and you'll probably get hired by an NHL team!

[Resolute 14]

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.

I was referring to the condescending nature of your post, which was basically "this is why you suck".

But to answer your question, it is actually baffling that you would take the position that the the standings are wrong as the basis for an argument. On paper, the Kings are more talented, yes. On paper, they should be better. But on the ice, the Flames have been the marginally better team so far this year, and that reflects in the standings.

The only thing the stats tell us is that it is highly unlikely that the Flames should be a better team this year, but that does not invalidate the standings themselves. The 1991-92 Flames should have been a team that went to the third round of the playoffs. Looked great on paper, but were flat out a bad team that year.

And if you want to get into arguments of "Luck", the Kings PDO is 99.9. (100.5 at home, 99.5 on the road). They aren't unlucky, just mediocre.

[Enoch Root]

You keep referring to your sig...

It is actually a great example of where so many of the analytics guys go off the rails - overconfidence.

The example is a pure random event - a coin flip. It has a 50% probability of either outcome. A grade 7 student could tell you that those odds are in your favour and that you should keep betting because the stats will eventually take care of you.

I will ignore the fact that some people seem to think this is wisdom, and move on to the more important issue: Corsi is nothing like a coin flip. Even though teams will oscillate around 50%, does not mean that being above 50% puts you in the driver's seat.

There are far too many other variables involved.

Even though the analytics supporters know that, they still talk like they have 'additional knowledge' and an inside track on understanding the game. Instead of talking about the analytics as one (incomplete) piece of information within the conversation, they immediately dismiss anyone who questions them as troglodytes.

When you use Corsi to determine which teams are better, you are NOT getting better than 50% odds on a coin flip. But the analytics crew act like they are. And that pisses people off.

If anyone would like a great read on the science of human nature and predictions, there is a great book by Daniel Gardner called 'Future Babble'. It does a great job of discussing why we feel the need to predict the unpredictable, and the various traps that the experts fall into when doing so. Overconfidence is, not surprisingly, one of the biggest.

[CorsiHockeyLeague]

That was a long way of providing a bald assertion that, all indication to the contrary (i.e. that chart posted on a prior page that had different stats' relationship to points earned) that you don't think possession statistics provide any predictive value. You're welcome to that view.

[squiggs96]

Originally Posted by Resolute 14

I was referring to the condescending nature of your post, which was basically "this is why you suck".

But to answer your question, it is actually baffling that you would take the position that the the standings are wrong as the basis for an argument. On paper, the Kings are more talented, yes. On paper, they should be better. But on the ice, the Flames have been the marginally better team so far this year, and that reflects in the standings.

The only thing the stats tell us is that it is highly unlikely that the Flames should be a better team this year, but that does not invalidate the standings themselves. The 1991-92 Flames should have been a team that went to the third round of the playoffs. Looked great on paper, but were flat out a bad team that year.

And if you want to get into arguments of "Luck", the Kings PDO is 99.9. (100.5 at home, 99.5 on the road). They aren't unlucky, just mediocre.

They should likely not even play NHL games anymore. Before the season each team assembles their roster. Then the advanced stats guys use their computers to tell us how the season should play out. The winning team is awarded the Cup in October and everyone is for the better.

I don't hate advanced stats, but I find the people advocating for them are trying so hard to make everyone believe what they feel is right. I think there is a place for advanced stats, but so often when something doesn't fit in the linear mode of thinking for an advanced stats guy, they go off on tirades about how everyone else is wrong/stupid/ignorant/delusional/etc. This isn't a debate over gravity or what shape the Earth is where there are hundreds of years of data to back this up. Corsi and Fenwick are a couple of years old, and whether they want to admit it or not, it's not a perfect system and not every kink has been worked out.

My two biggest complaints over advanced stats are 1) They people who advocate for them seldom want to hear an opposing view and 2) They use unreasonable timelines and "forget" factors to help prove their worth.

1) I think many of us were open to the idea of using advanced stats when they first started coming out. If a system will help me better enjoy or understand hockey, I'm all for it. I'm not sure advanced stats guys even like hockey. They just want to point and say "see I was right". When introducing the actual standings, team chemistry, injuries, etc. if it doesn't fit the advanced stats rhetoric, they won't even listen to you. That's ignorant.

2) The biggest statement they've made that discredits them is that Colorado is a bad team this year so we were right that they were a bad team last year even though they won their division. If Calgary has a bad season next year it proves we were right that they were a bad team this year. No it doesn't. If Calgary is bad next year it proves they are bad next year, especially with roster turnover.

[Resolute 14]

^In fairness, how many times in the Sutter years did we look at our roster and lament that we were better on paper than in real life? Analytics adds value, but their strongest proponents are no better than their strongest opponents for allowing their preconstructed belief to colour what they see in the numbers. So while I can certainly understand why CHL so frustrated at the negative reaction he and other proponents get, I also see the irony in how they don't realize they act the very same way. Two sides of the same coin.

[automaton 3]

Flip side of the Flames equation - if the advanced stats for the Kings are good, then why are they struggling in the standings? What factors are the current models missing? At this sample size, luck/randomness should not be much of a factor.

The advanced stats as they stand are mildly interesting, but they are only capturing a small portion of the game "events". Once better player and puck tracking systems are brought into play, the statistics should also become more useful.

[automaton 3]

Originally Posted by squiggs96

2) The biggest statement they've made that discredits them is that Colorado is a bad team this year so we were right that they were a bad team last year even though they won their division. If Calgary has a bad season next year it proves we were right that they were a bad team this year. No it doesn't. If Calgary is bad next year it proves they are bad next year, especially with roster turnover.

This one bothers me too. The Avs made significant lineup changes (losing their #1 centre in particular) so a drop off in performance should be expected.

Teams are dynamic and can't be easily compared year to year.

[calf]

The problem I have is the notion that advanced stats like Corsi are a predictive tool.

Maybe that's the case for fans and observers, but if I'm running a team, I'm using the stats as an analysis tool, and reconciling the gaps between what the stats indicate should happen vs what actually is. The second you rely on "the stats say we should be winning more than we are, so we'll keep doing what we're doing" is the second you're missing the point. Perhaps after looking at all the factors, you are just unlucky and it's best to stay the course. Or, after reviewing the tape you're finding your guys are strong on the puck, but have bad shooting accuracy and keep firing the puck right at the goalie's chest. Maybe a team has terrible possession stats, but keep the other team to the outside, and have the speed and skill to score on the rush, so nothing needs to change.

Basically, it operates under the assumption that the past predicts the future, without taking into account how teams learn from the past to change how they play (or not) in the future based on expected/actual results.

[CaptainCrunch]

I'd like to see your proof on your signature.

And does that calculation take into account random events, differences in tossing mechanics, effects of environment (table surface, air currents, some dude sneezing in your face.

Vegas gambles based on the theory above that even based on a linear betting strategy the player if he plays long enough reduces the odds of winning. If your signature was correct, Vegas wouldn't be in business.

Combine that if you calculate the probability of tossing lets say 65% losing and 35% winning over lets say 500 tosses, and then enact the tosses in real life then even if you end up not reaching the result of the calculation one or twice or 3 times out of 10 then the calculation is flawed.

The problem with advanced statistics is it makes the assumption that every team plays the same way, that every first line LW plays the same that every bottom 6 player plays the same over a set period of time, and explains a team winning not by systematic differences but by luck, which is the cop out formulation if you can't take into account inconsistencies int the environment full of inconsistencies.

If advanced statistics was the end all and be all then it could accurately predict results through a set time period (season, game period, etc) every time.

I would argue that advanced statistics is extremely valuable as a measuring tool to look at player performances in the past, but I would also argue that advanced stats, that can't take into account random events, inconsistencies between two teams in terms of game planning puck consistancies, quality of scoring chances etc, that its not a predictive tool.

I mean frankly and going back to your signature, Advanced Stats is a beautiful way to game a system, to induce people to bet, but in the end unless you are tossing coins in a vacuum with a nasa calibrated robot arm doing the tossing then its not something you can base predictions on.

A guy like Gaudreau for example is a fly in the ointment in terms of possession numbers, because he might go through a game with a very low possession number and still burn you with two goals.

Kris Russell by all advanced Stats the other night had a terrible game, however he was incredibly key in terms of winning games.

This isn't money ball, and thats the issue with Advanced Stats users, they believe that they can be predictive with hockey like with baseball. But baseball is a fixed and fairly linear game with specific roles. In hockey the minute that a defenseman for example makes a decision to not do a defenseman like thing, the stats don't support it.

The whole, unsustainable because of Colorado, and Toronto because look at advanced stats, is flawed, if Calgary had the exact same line up issues that those two teams had (Toronto had no defensive conciousness, got great goaltending and while they had good top pairing bluelines they didn't have much after that) Colorado had fabulous forward talent and AHL blueline and got great goaltending). Then you could argue that advanced stats means something.

but Calgary isn't winning the same way as those teams, certainly don't have the same team construction, and what looks like higher team fitness then any other team in the league (because the fitness level is another calculation that's not taken into account).

I'm a believer that statistics can help you analyze your talent to a point, but as far as predicting that the Flames aren't sustainable or they're going to drop badly next year without even knowing next years lineup, or game time strategy etc. Is just blustering without knowing.

By the way, if you're looking at flipping 3 heads f versus 7 tails is between 10 or 15 percent

[GioforPM]

I can't lose a coin flip but win more afterwards if I try harder or add a new coin, or use a new technique.

On the other hand, I can in hockey.

It's funny because some of the weaknesses in advanced stats come from using too small of a sample size. However, other problems occur in using stats from the beginning of the season, without recognizing that teams evolve. So crappy early season results influence today's stat picture (not an "advanced" stat, but if you look at the Flames' PK % you'd think they are a pretty mediocre PK team right now. But they aren't).

[Street Pharmacist]

Originally Posted by CorsiHockeyLeague

That was a long way of providing a bald assertion that, all indication to the contrary (i.e. that chart posted on a prior page that had different stats' relationship to points earned) that you don't think possession statistics provide any predictive value. You're welcome to that view.

No, that's not his "bald assertion".

Before we begin here, I'm with you. Most of the anger and ignorance here towards new statistics is due to emotion and lack of understanding.

What he is saying, is that while shot attempts tell part of the story and do provide some predictive value, the story doesn't even come close to beginning and ending there. That's is where Enoch's, Resolute's and my frustration is.

For example, McCurdy's playoff prediction chart. It arbitrarily puts Fenwick far, far, far ahead of win%, goal differential and other predictors. While statistically, Fenwick close may be a better predictor, it's only very, very MARGINALLY better. Why should so much emphasis be put on that and ignore: strength of schedule, win%, goal differential or other predictors when the difference is so marginal?

That's where the frustration is. Shot attempts are a very useful part of analysis. They are not analysis. When a writer like Lambert says shot attempts show Calgary is terrible and everything else is just luck, that's the kind of singular thinking that made Toronto pick up Clarkson. Maybe a wider view is useful here.

As I've said earlier, Toronto and Colorado mean nothing. Toronto plays worse by every metric near the end of the season, not just corsi. Why do they get so many wins at the beginning of the season? Lambert's type of "analysis" would suggest it's luck. Really? The exact same pattern three years in a row? That's a pretty unsupported assertion. Colorado made some pretty significant personnel changes. Why is "luck" the explanation for their regression this year? They were pretty damn close to getting to the second round last year, on "luck"?

The assertion consistently is: Corsi wins and loses games, anything that is variable is "luck", or randomness. Despite the fact that if Calgary was getting "lucky" this year, that many wins would be a generational happening, not a few teams a year.

Having good Metrics is essential to being elite and having success. Being terrible will give you bad corsi. That doesn't mean that corsi explains everything between .

[CSharp]

I agree with Lambert's (Puck Daddy's) assessment of the puck possession game that the Flames lack. Whatever negative stats you look at with the Flames, on any given season over the past 34 years or so, it's hard to see a Flames team win so many games when they're down and out by mid-3rd-periods. If you just read the post title for this thread alone and don't read further into what Lambert's comments, no, they're not screwed for now. But if the Flames don't address the issues that they have in regards to their size, defence, and puck control - yeah, they might have issues after this season. The thing that amazes me is how many blocked shots there are in a Flames game and Russell blocks like half of the shots while Bouma gets about 25-30% on PKs. I don't care how good the Flames goalies can be and how many blocked shots there are per game/period, they should be playing more in the other end of the ice.

I'm also thinking that this season is kind of odd that the Flames have placed so much better in the standings while teams like the Kings and the Sharks are placing themselves just enough to within reach. I betcha next season when the NHL has changed the draft lottery system and the draft won't be as good in 2016 that the standings won't be so surprising. So, are the Flames screwed? They may be...

[CorsiHockeyLeague]

Originally Posted by Street Pharmacist

No, that's not his "bald assertion".

What he is saying, is that while shot attempts tell part of the story and do provide some predictive value, the story doesn't even come close to beginning and ending there. That's is where Enoch's, Resolute's and my frustration is.

That is not what I gather from this.

Originally Posted by Enoch Root

When you use Corsi to determine which teams are better, you are NOT getting better than 50% odds on a coin flip. But the analytics crew act like they are. And that pisses people off.

What this suggests to me is that he thinks it has no predictive value. This says to me, "you're not improving your odds of being right in predicting future wins by looking at possession data." Unless I misunderstand.

The assertion consistently is: Corsi wins and loses games, anything that is variable is "luck", or randomness.

Like I say I'm tired of having this argument so I don't want to get into your whole post, but this is pretty obviously a straw man if you've been following the discussion as I know you have. Sure, a lot of it is randomness, which should be obvious if you've ever watched hockey. But there's plenty outside of shot attempts that win and lose hockey games - shooting percentage, save percentage, special teams percentages, penalties drawn or taken, CHIP, and so on. I don't think any reasonable read of what I've said in here, or any decent analytics writer's work, would accord with what you're attempting to foist on those writers as their view, consistent or otherwise.

[Street Pharmacist]

Originally Posted by CorsiHockeyLeague

That is not what I gather from this.

He's giving an example of misuse, not a comparable predictor.

What this suggests to me is that he thinks it has no predictive value. This says to me, "you're not improving your odds of being right in predicting future wins by looking at possession data." Unless I misunderstand.

I understood it as is value it's to the exclusion of other complexities

Like I say I'm tired of having this argument so I don't want to get into your whole post, but this is pretty obviously a straw man if you've been following the discussion as I know you have. Sure, a lot of it is randomness, which should be obvious if you've ever watched hockey. But there's plenty outside of shot attempts that win and lose hockey games - shooting percentage, save percentage, special teams percentages, penalties drawn or taken, CHIP, and so on. I don't think any reasonable read of what I've said in here, or any decent analytics writer's work, would accord with what you're attempting to foist on those writers as their view, consistent or otherwise.

I think you're lumping my argument with others. You have not said shot attempts the only value. I feel you've been quite neutral and fairly informative.

I'm only trying to provide context to the frustrations with "the analytics community" as it stands right now. I know at least myself and Resolute read a fair bit and see much value in this exercise.

Where I'm trying to draw your attention is to the main writings that get views: Lambert (who you've hinted at first didn't say anything inflammatory), Drance, Mirtle (occasionally), Yost, and a variety of others. In fact, not one have I read that the Flames must be doing something other than getting lucky. Every single article has described their deviance from the prediction as anything other than luck.

This assertion and the fallacy of pointing out Toronto and Colorado every time as some trump card are what gets our panties in a knot

[CorsiHockeyLeague]

I don't believe there are many if any analytics writers (or any who anyone listens to) who could be accused of taking the view that any success not resulting from possession is based on luck.

I don't view Lambert as an analytics writer. He does use some advanced stats to make points, but that's not really what he writes about. Guys like Drance (who barely writes about this stuff anymore) and Yost clearly are. Mirtle's a bit of a weird case because he values the statistics and uses them to supplement his views but isn't really an "analytics guy". Of these guys, only Yost is an analytics guy in the vein of Rob Vollman or Garrett Hohl or Tyler Dellow. The difference, for me, is that Mirtle and Lambert write opinion pieces that rely to some extent on stats commonly associated with analytics, as opposed to actually asking a general question (e.g. "how much impact does faceoff winning percentage have on the outcome of games?") and trying to answer it.

[Scornfire]

Agreed with SP, it's not you in particular, frankly you've made some extremely valid points. But the most vocal and prominent of the advanced stats crowd are the ones who imperiously flaunt their metrics as the end all be all and pass everything else off as luck and randomness. I mean honestly, have people like lambert ever played a bloody hockey game.

This one pretty much took the cake for me, Vancouver fan from the other boards

it's not. it's that unsustainable play has no preordained start and end time. play enough games of skill with a chance element (like poker or backgammon) and these things will become very clear. you will see luck last a lot longer than it has any right to, and you will see others take bad beat after bad beat after bad beat (or joker after joker) no matter how well they play.

Seriously equated poker to hockey, /sigh

[GioforPM]

Originally Posted by CorsiHockeyLeague

I don't believe there are many if any analytics writers (or any who anyone listens to) who could be accused of taking the view that any success not resulting from possession is based on luck.

I don't view Lambert as an analytics writer. He does use some advanced stats to make points, but that's not really what he writes about. Guys like Drance (who barely writes about this stuff anymore) and Yost clearly are. Mirtle's a bit of a weird case because he values the statistics and uses them to supplement his views but isn't really an "analytics guy". Of these guys, only Yost is an analytics guy in the vein of Rob Vollman or Garrett Hohl or Tyler Dellow. The difference, for me, is that Mirtle and Lambert write opinion pieces that rely to some extent on stats commonly associated with analytics, as opposed to actually asking a general question (e.g. "how much impact does faceoff winning percentage have on the outcome of games?") and trying to answer it.

So only the guys who use analytics in ways you agree with are analytics writers? "No True Scotsman...."

[SuperMatt18]

In Baseball advanced stats have been around longer, and they still don't predict anything at the team level. Good at looking at individuals but ignores a lot of other variables when building a team.

Look at the last couple years, the Oakland A's have still yet to win a world series. Yet a team that seems to spit in the face of all the advanced stats has won 3 out of 5 world series, on what advanced stats call "luck".

IMO Advanced stats are a great tool, they show you some really interesting metrics, and can be great at comparing individual performance but they have some flaws at the team level.

I think a couple of issues with advanced stats are that they discount elite goaltending (it's not "luck" that Nashville, Montreal, and New York have high PDO) and style of play.

Shot blocking is a legitimate strategy. A smaller, speedier team like the Flames will likely be a poor possession since they create more off the rush and struggle in the defensive zone at containing bigger forwards leading to possession in the zone against them.

The other issue is that it is really only good at evaluating team play in-season, roster turnover happens at such a high rate that it's not going to predict things year to year for teams.

Let's look at Corsi Close from last year and what it "predicted this year"

West Top 8 (4/8):

LA
Chicago
San Jose
St. Louis
Vancouver
Dallas
Anaheim
Arizona

East Top 8 (4/8):

Boston
New Jersey
New York
Ottawa
Tampa Bay
Detroit
Columbus
Florida

So last year's Corsi numbers seemed to accurately predict who would be good this year at a 50% clip.