[CP Story] The Calgary Ava-Leafs

[Bingo]

Originally Posted by EldrickOnIce

Regarding PDO, Flames goaltending is only statistically average. Have they stole a few? Absolutely. That's what you need them to do.
Shooting % is more interesting. They are going to have to shoot more, or goals for per game will drop... unless they really are elite offensively. I'm expecting/hoping it meets somewhere in the middle.

Agreed ... Hudler is off the charts and has to regress. But after that there are reasons for non concern.

Bouma? Yeah he'll come off but then he isn't expected to score. 5 goals this far is likely his season, but then he's taken a bigger role with guys that do score being injured

Jooris see Bouma

Raymond ... all early and he has nothing to do with the last 20 games

Wideman is likely the last of the eye popping concerns.

[EldrickOnIce]

Originally Posted by Enoch Root

Great article.

One question I would have is: sure, we all know about the Leafs and the Avs, but were there other teams - and if so, how many? - that also had similar stats but didn't regress?

We always here about the events that support our position/theory/prediction, but what about the events that didn't support it?

You brought up one.
The Avs made the playoffs. They are the proof.
Apparently regression waited til this year...

[getbak]

Originally Posted by GioforPM

In the last 10 games they've outshot the other team 5 times, been outshot 4 and tied one.

Their average is only a -2.2 shots, IIRC. I suspect if you take out the Chicago outlier (and let's say, to be fair, drop the Montreal loss as an example going the other way), the numbers get a lot closer.

I think the fact that the Flames are the top shot-blocking team in the League really skews their Corsi numbers. So, their actual shots are good, but their "advanced" numbers aren't.

Looking at Saturday's game:

Shots: Flames - 29 / Coyotes - 26
Shot Attempts Missed: Flames - 7 / Coyotes - 11
Shot Attempts Blocked: Flames - 16 / Coyotes - 26

The Flames led in shots by 3, trailed the "Fenwick" stat by 1, and trailed the "Corsi" stat by 11.


I can't imagine anyone who actually watched Saturday's game would suggest that Arizona carried the play other than briefly at the end of the second period.

[Phil Russell]

I asked this in another thread last week but never saw/may have missed an answer. Has anyone ever actually tested these advanced stats? Has a causal relationship been established between Corsi etc. and winning? Or, is everyone just assuming that Corsi etc. are measures of winning?

The San Jose game was the kicker for me. A 2-0 dominating win by Calgary. San Jose never dominating the play at any point in the game. And yet, San Jose won the advanced stats game. My guess is that advanced stats don't measure team success, or at the very least that there is a lot more to team success than can be measured by advanced stats. Again, however, this is something that a stats pro could test.

[Bandwagon Surfer]

Originally Posted by Hemi-Cuda

I hope the Flames ride their unsustainability all the way to the playoffs and all these bloggers that can't see past the new fancy stats have to eat a little crow

Unfortunatly, I do not think that will be enough to quiet the nervous voice in my head.

I want a cup, and to get a cup I want a team that can be a sustained cup threat for many years. So to really satisfy my doubts it will take seeing a year to year improvement in this team. Hopefully the doubting thoughts will leave me sooner, but realistically I think it will take two consecutive playoff runs, with the second run showing improvement over the first.

[Camronius]

Originally Posted by Phil Russell

I asked this in another thread last week but never saw/may have missed an answer. Has anyone ever actually tested these advanced stats? Has a causal relationship been established between Corsi etc. and winning? Or, is everyone just assuming that Corsi etc. are measures of winning?

The San Jose game was the kicker for me. A 2-0 dominating win by Calgary. San Jose never dominating the play at any point in the game. And yet, San Jose won the advanced stats game. My guess is that advanced stats don't measure team success, or at the very least that there is a lot more to team success than can be measured by advanced stats. Again, however, this is something that a stats pro could test.

Further to this I'm curious if anyone has gone back with advanced stats historically to look at teams past. That might shed some light if some teams defy the odds, how often the stats are right etc..

[Street Pharmacist]

Great timing Bingo.

I was crazy busy with a pile of things I needed to do at work, so I did what comes naturally - I procrastinated and looked further into corsi vs goal differential. The results were very surprising to me. Before I get to the results I'll explain what I did. Skip to the end if you don't want to read it, the good stuff is at the end.

I wanted to ask the question: Which metric is more reliable: corsi for % at 5v5, or total goal differential?

I don't care about goal differential and corsi at the end of the season. It's irrelevant in our discussion and that is what's missing here: time frame. End of season stats aren't predictive of what will happen, just descriptive of what did happen. If we want probabilities and predictions, we have to use the same time frames and compare the final results in my opinion.

I decided I would use October 1st to December 1st as that is where we are now. The sample size was limited by the fact that the NHL did not keep track of missed shots or blocked shots before 2005/06, so corsi isn't available before then. Secondly, there was no hockey between October 1st and December 1st in 2012 due to the lockout, so that season is eliminated from the data. That leaves us with 8 years of data, which is fairly robust.

The parameters for the Metrics were somewhat more arbitrary. I used corsi for% <45% as that was Drance's metric. I also used a goal differential of =\> +10. This was simply because it's a round number and the flames are above it. This does lead to some holes in the results, but we can discuss that later if you wish.



The results:

In the eight years tallied there were only 14 teams out of the 240 that had a corsi for% 5v5 before 45% (only 4 were worse than the Flames current corsi) as of December 1st. Of those 14, I was surprised to find that 4 of them made the playoffs. That's 28%!! That's quite concerning for folks who ascribe to the predictive power of corsi. To me, the fact that these 14 are the worst of the worst at corsi. The bottom 6%. Yet, over a quarter of them still made the playoffs. Clearly corsi is important, but maybe early in the season it isn't that predictive.

The goal differential was equally interesting. There were 62 teams with a total goal differential over +10 by December 1st. Of those, only 6 missed the playoffs. That's 9.7%. To add more optimism for flames fans, no team since 09/10 that was above +10 has missed the playoffs. To add a little sting, the last team to do it was the flames (and Hartley's Atlanta).


In the end, there's legitimate reasons to be concerned, and legitimate reasons to be optimistic. To me, the fun is that the games still have to be played and tons can change. I enjoy this process of digging up stats and looking for road signs to know where the team might be headed, but nobody can speak with as much certainty as a few bloggers seem to be.

After looking at it this way I've become even more optimistic about the flames this year!

[Street Pharmacist]

Originally Posted by Phil Russell

I asked this in another thread last week but never saw/may have missed an answer. Has anyone ever actually tested these advanced stats? Has a causal relationship been established between Corsi etc. and winning? Or, is everyone just assuming that Corsi etc. are measures of winning?

The San Jose game was the kicker for me. A 2-0 dominating win by Calgary. San Jose never dominating the play at any point in the game. And yet, San Jose won the advanced stats game. My guess is that advanced stats don't measure team success, or at the very least that there is a lot more to team success than can be measured by advanced stats. Again, however, this is something that a stats pro could test.

They have been compared with real results ad nauseum and they do correlate with success.

We need to stop calling them "advanced stats". They're not advanced, they're just stats. In fact, corsi is actually simpler than the shots at use now. Corsi measures all picks fired towards the net. SOG measures only those spots that the goalie saves or goes in.

Do we question shots on goal? Faceoffs? Powerplay? Do they correlate with success?

[PeteMoss]

Bottom line for the Flames - someone needs to determine what their baseline shooting percentage is.

It is very high right now. It will come down in all likelihood. But how far will it come down. Since Hartley has been the Flames coach - they have a top half shooting percentage each year. Based on that I would expect the Flames to end up high... I wouldn't expect to see an ultra drop off.

Also worth noting - the Flames don't have the highest shooting percentage in the league. They are behind Tampa. And there are four teams above 11% right now - Tampa, Calgary, Pittsburgh and Toronto. Its not like they are some crazy outlier.

[Enoch Root]

Originally Posted by PeteMoss

Bottom line for the Flames - someone needs to determine what their baseline shooting percentage is.

It is very high right now. It will come down in all likelihood. But how far will it come down. Since Hartley has been the Flames coach - they have a top half shooting percentage each year. Based on that I would expect the Flames to end up high... I wouldn't expect to see an ultra drop off.

Also worth noting - the Flames don't have the highest shooting percentage in the league. They are behind Tampa. And there are four teams above 11% right now - Tampa, Calgary, Pittsburgh and Toronto. Its not like they are some crazy outlier.

This is an interesting point.

It is important to keep in mind that offense ebbs and flows, that the game constantly changes, and that there is an ongoing tug of war between offense and defense - one will get the upper hand in some way and then the other will adjust.

So the question should be asked: is it simply the case that shooting percentages are unusually high this year for some teams? A random event that will normalize at some point?

Or is something else going on? Are offenses starting to discover a new way to be successful? Is there a new offensive trend that creates more quality chances relative to total shots taken?

The 4 teams listed above are all fast skating teams that generate a large amount of their offense off the rush. Does that result in a higher goals/shots ratio? Are these teams adapting in a way that affects their shooting percentages?

I think it is at least a possibility and a question that needs to be asked.

[Itse]

Originally Posted by Street Pharmacist

I wanted to ask the question: Which metric is more reliable: corsi for % at 5v5, or total goal differential?

Great job, thanks. That does say pretty conclusively that there is no reason to be overly worried about corsi, as long as goal differential is good.

That said, you have to remember that this is not an either/or question. You can use both at the same time. My question would be, do predictions get notably better by taking corsi into account, and if so, how much?

So for example, if you only look at teams with a good +/-, are the ones that drop out of playoffs more likely to have bad corsi?

And of course, are the teams with a poor goal differential that eventually make the playoffs more likely to have good corsi?

(Too lazy to do the math myself, sorry.)

[Textcritic]

Originally Posted by Bingo

Agreed ... Hudler is off the charts and has to regress. But after that there are reasons for non concern.

Bouma? Yeah he'll come off but then he isn't expected to score. 5 goals this far is likely his season, but then he's taken a bigger role with guys that do score being injured

Jooris see Bouma

Raymond ... all early and he has nothing to do with the last 20 games

Wideman is likely the last of the eye popping concerns.

While all these players are likely to see their shooting percentages shrink, we should also expect a few players to see theirs show notable increases. In particular, I think it is a good bet that Gaudreau (9.1), Glencross(9.3), and Colborne (0.0!) are all likely to see dramatic improvements in their individual shooting percentages as the season wears on.

[Textcritic]

Originally Posted by Enoch Root

...Or is something else going on? Are offenses starting to discover a new way to be successful? Is there a new offensive trend that creates more quality chances relative to total shots taken?

The 4 teams listed above are all fast skating teams that generate a large amount of their offense off the rush. Does that result in a higher goals/shots ratio? Are these teams adapting in a way that affects their shooting percentages?

I would hazard to guess that this is probably the case. It makes some sense to me that teams with average or below average footspeed would need more time, more space, and more shots to generate goals. Are cycle teams also teams that have a limited number of players with above average skating skills and speed? That is the one thing that really stands out in virtually every Flames game I have seen this season—they are FAST. Really, really fast. Perhaps more importantly, they have a good number of fast players who can also make plays at top speed and finish them.

[Phil Russell]

Originally Posted by Street Pharmacist

They have been compared with real results ad nauseum and they do correlate with success.

We need to stop calling them "advanced stats". They're not advanced, they're just stats. In fact, corsi is actually simpler than the shots at use now. Corsi measures all picks fired towards the net. SOG measures only those spots that the goalie saves or goes in.

Do we question shots on goal? Faceoffs? Powerplay? Do they correlate with success?

Thanks for answering my question about advanced stats. I suppose if everyone started latching on to faceoff percentage as the be all and end all measure of team success then yes we would be questioning if faceoffs correlate with success.

[EldrickOnIce]

Originally Posted by Textcritic

I would hazard to guess that this is probably the case. It makes some sense to me that teams with average or below average footspeed would need more time, more space, and more shots to generate goals. Are cycle teams also teams that have a limited number of players with above average skating skills and speed? That is the one thing that really stands out in virtually every Flames game I have seen this season—they are FAST. Really, really fast. Perhaps more importantly, they have a good number of fast players who can also make plays at top speed and finish them.

And then you have teams (or players) like the Oilers, who appear to be playing for corsi. 1 on 3 rushes where you drift a harmless shot at goal. You do that a couple, three times each period and you might bump corsi 5 points or more.
Makes Dellow happy, but not conducive to winning.

[PeteMoss]

Originally Posted by EldrickOnIce

And then you have teams (or players) like the Oilers, who appear to be playing for corsi. 1 on 3 rushes where you drift a harmless shot at goal. You do that a couple, three times each period and you might bump corsi 5 points or more.
Makes Dellow happy, but not conducive to winning.

Also - always try to carry the puck in. Carrying the puck in = GOALS. Even if you are 1 on 3 and everyone else on your team is heading off for a line change so when you turn it over its a 2 on 1 going the other way.

[EldrickOnIce]

Originally Posted by PeteMoss

Also - always try to carry the puck in. Carrying the puck in = GOALS. Even if you are 1 on 3 and everyone else on your team is heading off for a line change so when you turn it over its a 2 on 1 going the other way.

Absolutely.
Just talking from a purely statistical standpoint. You can introduce bias to any data, and therefore render skewed results.
The Oilers had an organizational goal to increase 'possession numbers'. They hired Dellow. The players themselves have talked about wanting to improve possession.
So you create the data point events, and not the other way around. They prove you can artificially raise possession numbers while playing no good.

[Tinordi]

Originally Posted by EldrickOnIce

Absolutely.
Just talking from a purely statistical standpoint. You can introduce bias to any data, and therefore render skewed results.
The Oilers had an organizational goal to increase 'possession numbers'. They hired Dellow. The players themselves have talked about wanting to improve possession.
So you create the data point events, and not the other way around. They prove you can artificially raise possession numbers while playing no good.

It's not that simple. Infact there are few blueprints for ways for GMs to improve their possession. Do you go out and acquire high corsi-rel players? Maybe. But it doesn't appear like that's been the strategy employed in Edmonton. Tyler Dellow probably would have been the first guy to ask why they only had 2 NHL centers? That will impact team performance way more than their corsi score.

It's cherry picking to say that this team prioritized 'corsi' and this team is team is now failing therefore corsi is no good.

What these inquiries do is build insight in the game and probabilities related to success. You're taking an anti-statistical view using one anecdote in a larger panoply of analysis.

Anyway, back to the article. Great stuff. We're riding a great shooting percentage. Sometimes that's luck, that's confidence, swagger, coaching. But it really tends to not sustain itself over the long term. Toronto also had really good shooting percentages when they went to the playoffs two years ago. It's since fallen back to the average. Back then, Leafs fans would argue until they were blue in the face that the high percentage was from coaching and skill Similar to where we are now as Flames fans.

[EldrickOnIce]

^I'm not saying corsi is no good.
I'm saying it is that easy to skew the data, if you are playing to create corsi events instead of creating scoring chances.
And I think the Oilers are doing that, to some degree.
because they are no good

[PeteMoss]

Originally Posted by Tinordi

It's not that simple. Infact there are few blueprints for ways for GMs to improve their possession. Do you go out and acquire high corsi-rel players? Maybe. But it doesn't appear like that's been the strategy employed in Edmonton. Tyler Dellow probably would have been the first guy to ask why they only had 2 NHL centers? That will impact team performance way more than their corsi score.

It's cherry picking to say that this team prioritized 'corsi' and this team is team is now failing therefore corsi is no good.

What these inquiries do is build insight in the game and probabilities related to success. You're taking an anti-statistical view using one anecdote in a larger panoply of analysis.

Anyway, back to the article. Great stuff. We're riding a great shooting percentage. Sometimes that's luck, that's confidence, swagger, coaching. But it really tends to not sustain itself over the long term. Toronto also had really good shooting percentages when they went to the playoffs two years ago. It's since fallen back to the average. Back then, Leafs fans would argue until they were blue in the face that the high percentage was from coaching and skill Similar to where we are now as Flames fans.

Leafs shooting percentage:

2014-15 - 4th in the league - 11%
2013-14 - 6th in the league - 9.69%
2012-13 - 1st in the league - 11.47%
2011-12 - 6th in the league - 9.78%

Looks to me like it is a skill they have.