[PGT] Coyotes 4 Flames 1

[Saqe]

You also create your own luck by showing up every night.

Didn't see that happen so much this year.

[Cube Inmate]

Originally Posted by Bingo

The level of disdain for simple counting stats on this site is mind blowing, and I hope the Flames don't fall into either of the following traps when the season is over.

a) go only on advanced stats and assume everything is fine
b) assume the counting stats are wrong and just blow things up (coach, key players)

With this season's Flames, the counting stats you refer to are simply not correlated to the results in the way you think they should be. That leaves you with one of two options: you can conclude that the causal relationship that you hypothesized is wrong, or you can blame "bad luck."

I've said repeatedly that "bad luck" isn't a thing. It's an excuse used when the reality of what occurs doesn't line up with our subjective impression of what "should" occur. Maybe you can excuse the occasional bad bounce, freak injury, or bad call on misfortune, but leading the league in missed shots has zero to do with luck. It's simply the result of factors that you obviously haven't correctly identified yet.

So at this point, we know the theory behind the statistical impact is broken. High number of shots taken doesn't, by itself, correlate to winning. So-called "high-danger" scoring chances aren't actually high-danger, since they aren't leading to a higher number of goals. Your theories are broken.

But what you're advocating is taking the data behind these broken models and trying to find some new meaning for it. Well, you'll always be able to find some correlation to existing data that fits your agenda -- that's the nature of correlation. But it would be a waste of time doing so unless you have a reason to believe that correlation (causation, actually) should exist.

Flames this year are bad.
We thought "lots of shots on net" would lead to winning.
We were wrong.

Management needs to find a new theory, test that theory against proven winners (i.e. regression analysis), and start again next year. Poring over data based on this year's team is pointless.

[Bingo]

Originally Posted by Cube Inmate

But what you're advocating is taking the data behind these broken models and trying to find some new meaning for it. Well, you'll always be able to find some correlation to existing data that fits your agenda -- that's the nature of correlation. But it would be a waste of time doing so unless you have a reason to believe that correlation (causation, actually) should exist.

Actually if you read what I said you'd see I'm advocating looking at everything, digging into the stats and seeing if you can find a reason that what looks to be good or high may not be.

Not sure why you'd boil it down to a flustered scientist want to make his hypothesis true when it isn't.

[DeluxeMoustache]

Originally Posted by Bingo

Well then it's a one year sickness of being bad at shooting the puck because they didn't have that problem last year.

Luck doesn’t work like that. Consistently bad luck for a full season across the team? Please.

There is another narrative, with many facets that clearly make at least as much sense as dumb luck.

It starts with counting stats and isolating teams from league. Identifying Team philosophy and style of play impacts, and distinguishing how they may deviate from broad brush league average numbers.

14-15 - the direction in the D zone was giving the other team the outside, and a quick transition once you got the puck. (which is consistent with poor disgruntled Sarich lamenting the frowning on back passes). It was a reversal of Brent Sutter saying explicitly that they aren’t a team that was going to score off the rush.
A guy like Russell was skewered on counting stats but also successfully blocked more shots than anyone else in the league.
They played with speed, generated a lot of odd man rushes (which I consider a better indicator of high danger than bloody shot location) and had a high shooting percentage.

That all makes sense. It itemizes several elements that expose the weaknesses of counting stats.

They were not playing for Corsi, they were playing rope a dope and trying to catch the opponent flat footed. Very exciting offensively when they got the puck, but a roster that has to work hard defensively to recover it.

Yes, across the board, considering all teams and all systems, the models work, to an extent. The influence or contribution of individual systems is removed from the massive sample size. Of course, more shots in general, across the entire league, leads to more goals.

But when you have outliers, then you look at model assumptions.

Assumption that shot location is a proxy for danger?
Challenged by comparing cross crease tap in vs. stuffing a puck in to a goalie’s pad.
Challenged by comparing 2 on 1s with time and space to hurried forced shots where the goalie is set

Assumption that total shots should imply control of play and likelihood of outscoring opponents?
Challenged by idea that shot quality affects likelihood of scoring. Reinforced by 10 game stretch where the Flames outshoot their opponent 400-275 and are outscored by 16 goals (!). And go 3-6-1. Too consistent to be luck. Perhaps something systematic that contributes to the consistent defiance of a counting stats based model?

Assumption that all players on a team are unlucky for a prolonged period of time ?
Luck just doesn’t work like that.

Brodie and Bennett both instantly turned to duds under Gulutzan? Something stinks there.

There are observations about the team’s style of play, repeatedly, that lead to a larger number of shots that are less dangerous, by the eye test, than their opponents. With justifications involving odd man rushes for and against, intensity and focus of defenders (ex. Several of E Kane’s goals), etc.

Face it. You have watched many games where the Flames outshot their opponent, let’s say roughly 40-30, and lost handily. They never looked like they were going to win those games.

They are a crappily coached slow team, lacking intensity, focus and confidence, and luck is pretty darn far down the list on why they have been so bad.

[Calgary4LIfe]

Originally Posted by Bingo

Hartley's 14/15 team had a lot go it's way, there's little doubt.

This season felt a bit like the antitheses to that. I'm certainly not calling the whole thing luck, but it's not just systems too. Bad bounces, odd calls, and the one that really sticks out to me is leading the league in missing the net.

Missed shots now up to 1207, when the next closest team is 1120 is just odd as hell. Missing the net could be execution too, don't get me wrong, but it's certainly not structure.

Before the torches get lit, I'm not suggesting they retain Gulutzan, but I'd want the GM to look at some of the unlikeliness of some of the stats to repeat if everything was left as is as part of the analysis.

Too many players on this team are below their position's average in shooting percentage. Is that all on coaching?

I just wouldn't discount anything.

What went their way exactly?

They got 'middle of the road' goaltending - the counting stats say as much.

They had more injuries - and to key personnel too - then at any time under Gulutzan.

Take a look at the defence. 12 different players played D for the Flames that year. That's a lot of rotation.

Backlund and Stajan both missed significant time - that left Monahan to fend for himself essentially, as the Flames had a terrible time trying to line-match.

I argue that it was also never just a single season - it was also the 2nd half of the previous season that the Flames started to play a consistent game (wasn't that they year they set - or tied - the franchise record for most 1-goal games?).

That season's Flames in my opinion were the antithesis to this year's Flames - poor (but sustainable) advanced metrics. This is where I argue that until advanced metrics starts to differentiate between quality of shots better (as in forcing a goalie to move in order to make a save, cross-crease passes, etc..) then it is 'additional information', not something you can call 'luck' (either good or bad) when the results don't align with the stats. It is too long of a sample size in both directions that have a negative relationship with the expected outcome to just merely call it luck.

Advanced metrics isn't 'garbage'. It just doesn't go nearly far enough. Hartley's system was geared towards allowing more shots from the perimeter and contesting the middle of the ice more, and blocking everything when possible, and then trying to create odd-man rushes. Odd-man rushes usually result in higher chances of scoring (I would imagine that goalies have to think about cross-ice passes more).

Gulutzan's system seems much more geared to smothering shot attempts in the defensive zone, and having a 5 man transition, and then trying to get the pucks into scoring areas. It results in less odd-man rushes, so one could reasonably conclude that with less odd-man rushes, the shot quality SHOULD be lower.

Does this make sense at all? I mean, this is just what I have been mulling over for a while now. Now which system is better? I think it is 100% team dependent. I think Hartley's system is far more effective for a quick team with puck-moving defencmen and young legs out there that can rush the puck. I think Gulutzan's system is probably better suited to an older team with defencemen that can't transition the puck as well (necessitating shorter passes/dump-ins) - but I think they need heavy wingers to attack the net for 2nd and 3rd chances MORE than what we have seen.

This is my take. I think coaching directly FOR CORSI is stupid. Goalies at this level might make an odd mistake and allow a goal in, but it is hard to beat an NHL level goalie even from a high danger area when the goalie is already set. Only elite players are capable of it, and even then a goalie SHOULD be saving most of them. I think that is why the Flames lead the league in missing the net as well, as they are forced to pick corners/5-hole as the opposing goalie is set.

This has been my take anyway, and I wish I could go back to the midpoint of the season and see if in fact those games that the Flames lost while vastly out-playing and out-chancing the opposition aligned with my theory (not making the goalie move enough, while having the opposition forcing the Flames' goalies move with their limited chances).

At any rate, going from Hartley to Gulutzan seems a heck of a lot like the Flames going from Keenan to Sutter. It was perhaps trying to make too much of an adjustment on play style.

[Bingo]

Originally Posted by DeluxeMoustache

Luck doesn’t work like that. Consistently bad luck for a full season across the team? Please.

Nice long lecture, but your summary is wrong.

I didn't say it was all luck. Not once.

I actually advocated digging into the high danger chance section to see what successful teams are doing differently when they run those stats up. I suggested three categories of high danger to see if you can ascertain why the Flames can add high danger events which have been proven to be 78% correlated to winning over 8 seasons, but still lose.

So thanks for the good stern talking to, but you missed your mark.

[Scroopy Noopers]

Originally Posted by Calgary4LIfe

What went their way exactly?

Immediately following that season Treliving (or Burke... can’t find the quote) said himself about making the playoffs:
“It’s like we won the masters, but had to sink 18 50’ putts to do it”.

[Resolute 14]

The level of disdain is pretty easy to figure out, I'd say. Too many people treat things like corsi as the One True Stat that is infallable. They seem to forget that the stat isn't even a measure of possession or quality. It is a measure of workload for a goaltender.

But we extrapolate and we assume, and while we do find a generally stronger correlation to success than many other statistics, that correlation is not nearly as strong as hardcore advanced stat types want to believe. And, as I see on Twitter all the damn time, these types immediately martyr themselves when people express skepticism.

I also find that, typically, a high corsi team doing well or a low corsi team doing poorly is treated as a validation, while a high corsi team doing poorly and vice versa is excused away. Most often as "luck". That creates a scenario where it becomes obvious people are only using the stat to confirm pre-existing conclusions. And that damages the overall credibility.

[Bingo]

Originally Posted by Resolute 14

The level of disdain is pretty easy to figure out, I'd say. Too many people treat things like corsi as the One True Stat that is infallable. They seem to forget that the stat isn't even a measure of possession or quality. It is a measure of workload for a goaltender.

But we extrapolate and we assume, and while we do find a generally stronger correlation to success than many other statistics, that correlation is not nearly as strong as hardcore advanced stat types want to believe. And, as I see on Twitter all the damn time, these types immediately martyr themselves when people express skepticism.

I also find that, typically, a high corsi team doing well or a low corsi team doing poorly is treated as a validation, while a high corsi team doing poorly and vice versa is excused away. Most often as "luck". That creates a scenario where it becomes obvious people are only using the stat to confirm pre-existing conclusions. And that damages the overall credibility.

I guess it's asking too much to read what I said huh?

Have suggested once and reiterated three times that I'd ask for a study of scoring chances looking for why the Flames are a top team in chances, but not a top team in scoring.

Is it a system issue?
Is it personnel?
Is it said personnel not performing?
Is it luck?

Yet over and over again the numbers are attacked and the point ignored.

Getting pretty tiring.

[delayedreflex]

Originally Posted by Bingo

I guess it's asking too much to read what I said huh?

Have suggested once and reiterated three times that I'd ask for a study of scoring chances looking for why the Flames are a top team in chances, but not a top team in scoring.

Is it a system issue?
Is it personnel?
Is it said personnel not performing?
Is it luck?

Yet over and over again the numbers are attacked and the point ignored.

Getting pretty tiring.

I found this older article (also part 1 here)which shows that shot location is not a good enough determiner of shot quality. Are the "high danger chances" that are usually tracked just based on shot location, or does it also take into account puck movement prior to the shot? According to this article, goalies still save like 90% of clean shots in the high danger area, even though overall they only save 75-80% of the total shots from that area. On transition shots and deflections, goalies seem to only save ~70% of the shots, and ~76% for shots off rebounds (though the datasets show a lot of variance due to the smaller sample size examined).

A key snippet from the article:

But by focusing on just the shot type and not the location, it became quite clear that scoring success in the NHL was predicated on movement and deception. If you remove these elements even the worst goaltender in the league begins to look like Dominik Hasek.

It would be amazing if we could see this sort of breakdown for the Flames this year vs. past years as well as vs. other teams. Many people here say the eye test says that the Flames aren't getting real high danger chances, ie. the type that actually has higher chance of scoring, vs. clean shots from the slot (which might show up as as "high danger" chance if shot location is the only thing tracked).

Understanding if Calgary's large number of "high danger" chances this year are real or not will help us understand if it's a matter of luck or not. I would say if Calgary is really generating high numbers of real high danger chances (shots off transition, deflections and rebounds), then that would point more to a luck and/or player execution issue - however if Calgary actually ISN'T generating lots of real high danger chances, and is instead just generating clean shots in the slot, then that points more towards a systems issue.

[Resolute 14]

I was speaking in generalities, Bingo. I was also noting that it's not just "this site".

[flamesfan1297]

It's probably been posted here already, but I seen it confirmed today that Quenneville and Bowman will return to the Blackhawks next year