"Sustainable" - Why use the Avs and Leafs as the parallel?

[Robbob]

Originally Posted by Resolute 14

In 2008-09, the Hawks were third with a CF% of 55.2, 15th with a PDO of 99.8 (7.41 sh% - 21st and .924 sv%, 7th). Their success was in their stingyness, they had the fewest Corsi events against with 2698. All of that at 5 on 5. Their Corsi close was fifth at 54.5. Interestingly, the Flames were second in CF% in both 5 on 5 and 5 on 5 close. In 2007-08, the Hawks were middle of the pack.

Yeah I just finished looking up some stuff as I wanted to see the progression of that team. Yeah, in the 07-08 year they were in the middle (except their SH%). CF% was 49.7 (16th), FF% 49.6 (18th), SH% 8.92 (5th), and SV% 91.85 (17th). I could not find anything of the 06-07 year.

[Calgary4LIfe]

Can someone please explain to an ignoramus like me PDO?

http://www.sportingcharts.com/nhl/st...rcentage/2014/

PDO is an advanced statistic in hockey that combines save % and shooting % in an effort to measure overall performance of a team and highlight teams with outsized "luck".
The basic idea is that teams on average should be around the 1.000 level as each shot is either a goal or a save teams that are relatively higher are producing above expectations and should regress back to the 1.000 level and teams that are relatively lower are producing below expectations and should regess up to the 1.000 level.

Honestly, this doesn't make sense to me. Why is shooting percentage and save percentage tied together in a metric that should equal 100? Anything above this is 'good luck' and everything below that is 'bad luck'?


Could it not be possible that having good players and/or a good system increase both metrics, without any luck involved? I'm sorry, but if I have a team with Carey Price in net, Ovechkin on a wing, Stamkos in the middle, and a guy like Subban blasting bombs from the point I would without a doubt expect my "PDO" to be higher.


Let's see the top 3 teams for this metric:


Montreal
Tampa

New York


Well, that makes fairly good sense, given the composition of the teams, no? Sure, Calgary at 5 and Ottawa at 6 may be the two outliers in the top 10, but again, to me it proves nothing.


Let's look at the bottom 3 teams:


Arizona
Edmonton
Carolina


Yep, makes sense there, given their performances and team makeups. You might argue that having Eric Staal, Jeff Skinner and even a surging Elias Lindholm may be a surprise, but the Canes have been terrible at scoring goals, and their save percentage is mediocre at best since all their goalies haven't been that great, and their defence (especially post-deadline) has been terrible.


It seems like a completely fabricated metric. Why should it equal 1.00? I understand they are trying to quantify luck here, but this is just stupid to me. Sounds a lot like the RGI metric here!



Also, CORSI - sure, some stuff is predictive and interesting, but it doesn't really measure possession now, does it? If you are saying you are measuring possession, then measure possession. All you are measuring is shots. Nothing more, nothing less.



Does the number of shot attempts correlate with possession? Sure it does, but not always! There are teams that will do their best at just getting shots to the net and hoping for the best, and others that try to make more plays. Surprise surprise - getting a tonne of shots to the net does NOT correlate with scoring chances, and it will lower your PDO (at least the shooting percentage side of things) to the 'average'.


I think there are some interesting things to take out of CORSI, but it is overblown and overused in my opinion. There are too many exceptions. PDO is only interesting as a gauge to see how good your players are - nothing more, nothing less (IMO).

[Quincy Egg]

Originally Posted by pylon

I really wish advanced stats would just get launched right into the sun.

You will never be able to generate a stat that measures team chemistry and work ethic. Advanced stat 'gurus' write every team that bucks the trend as an anomaly, and every team that happens to fall into the framework of the stats as a proven example. Duh, that's exactly how life is. Sometimes there is exceptions to the rule for unquantifiable reasons.

If advanced stats applied in real life, then only University educated people with IQ's over 130 should be wealthy. Well there are thousands of high school drop outs that buck that trend and become highly successful entrepreneurs because of work ethic and desire.

The stats only tell a tiny part of the story, and I truly believe some of these advanced stats guys, don't even like hockey, they just get a giant boner staring at an excel spread sheet, look at soulless numbers preying something in them confirms they are right, so they can froth at the mouth smashing the words "I told you so!" into their keyboards. They might as well just build paper rosters, and cheer for the piece of paper, and the piece of paper with the best calculation is awarded the Stanley Cup. That's what they seem to want.

This post just reeks of ignorance. Ever wonder why every single NHL team employs some form of advanced stats in player evaluation?

This isn't the mid 80s.

[Calgary4LIfe]

Originally Posted by Quincy Egg

This post just reeks of ignorance. Ever wonder why every single NHL team employs some form of advanced stats in player evaluation?

This isn't the mid 80s.

I do see his point though. It isn't just ignorance, and there is some truth to it.

I thought Burke was a bit of a dinosaur and a blowhard when he keeps knocking the advanced stats down. However, what exactly is he knocking down? Advanced stats as a whole, or the current metrics that are used?

Keep in mind that although every team does indeed employ advanced stats in player evaluations, what exactly are they? There are numerous teams - Chicago for sure, Flames (Treliving stated as much in an interview in the past), and probably most if not all - that have their own in-house stats that they value. Are they looking up CORSI, Fenwick, etc? Not sure if they do, but from the sounds of it, they sure place a high degree of emphasis on their undisclosed statistical information.

[Resolute 14]

Originally Posted by Calgary4LIfe

Can someone please explain to an ignoramus like me PDO?

...

Honestly, this doesn't make sense to me. Why is shooting percentage and save percentage tied together in a metric that should equal 100? Anything above this is 'good luck' and everything below that is 'bad luck'?

Well, since shooting percentage and save percentage are reciprocals, that is why the view is that the mean average would be 100 - it has to be. The "luck" part comes from the theory that given a large enough sample size, every team will trend (regress) toward that mean. Logic tells us that good teams will trend toward a PDO over 100, and bad teams below - exactly as you noted.

IMO, the argument that PDO is a proxy for luck is incredibly flimsy, especially relative to the argument that shot attempts are a proxy for possession. The Canadiens have a high PDO because Carey Price is a damn good goalie. The Flames have a high PDO because they are shooting the lights out. What we are doing may not be sustainable into next year, but that does not automatically mean we are lucky. It just means that we are bucking the trend.

[dobbles]

Originally Posted by pylon

I really wish advanced stats would just get launched right into the sun.

You will never be able to generate a stat that measures team chemistry and work ethic. Advanced stat 'gurus' write every team that bucks the trend as an anomaly, and every team that happens to fall into the framework of the stats as a proven example. Duh, that's exactly how life is. Sometimes there is exceptions to the rule for unquantifiable reasons.

If advanced stats applied in real life, then only University educated people with IQ's over 130 should be wealthy. Well there are thousands of high school drop outs that buck that trend and become highly successful entrepreneurs because of work ethic and desire.

The stats only tell a tiny part of the story, and I truly believe some of these advanced stats guys, don't even like hockey, they just get a giant boner staring at an excel spread sheet, look at soulless numbers preying something in them confirms they are right, so they can froth at the mouth smashing the words "I told you so!" into their keyboards. They might as well just build paper rosters, and cheer for the piece of paper, and the piece of paper with the best calculation is awarded the Stanley Cup. That's what they seem to want.

the bolded part reminded me of a story i read on 538 the other day.

http://fivethirtyeight.com/datalab/w...yball-metrics/

Neil, my favorite little tidbit from the the MIT Sloan Sports Analytics Conference was the fact that each day, all the stats nerds streamed into the building toward panel discussions about sports analytics, waltzing right past — actual sports! There was a massive regional volleyball tournament taking place in the same building, and as far as I could tell, no one stopped to check it out. And, honestly, part of me thinks that it could have been an NFL game and people would have still hustled upstairs to talk about sports rather than just watch them. There was a panel discussion called “Is Analytics Taking the Fun Out of Sports.” This felt like a nice parallel to that idea.

[mhsyyc]

I dunno but it seems to me most advanced stats guys are a-holes. They should just lighten up some and enjoy the game.

[BACKCHECK!!!]

Originally Posted by pylon

If advanced stats applied in real life, then only University educated people with IQ's over 130 should be wealthy. Well there are thousands of high school drop outs that buck that trend and become highly successful entrepreneurs because of work ethic and desire.

Right, and the people arguing against advanced stats are like the losers who say "Hey, even Bill Gates dropped out of high school, school is just holding me back, man."

The fact is that if you are actually making a plan which you are going to follow, you're better off with a plan that involves a good education, rather than a plan that involves dropping out of high school and "bucking the trend".

[Meers]

Originally Posted by BACKCHECK!!!

Right, and the people arguing against advanced stats are like the losers who say "Hey, even Bill Gates dropped out of high school, school is just holding me back, man."

The fact is that if you are actually making a plan which you are going to follow, you're better off with a plan that involves a good education, rather than a plan that involves dropping out of high school and "bucking the trend".

Actually, Bill Gates didn't drop out of high school. He dropped out of Harvard. A slight difference. It may even go further to proving your point.

[the2bears]

Originally Posted by Quincy Egg

This post just reeks of ignorance. Ever wonder why every single NHL team employs some form of advanced stats in player evaluation?

This isn't the mid 80s.

Tell us again how you were right about Colorado.

[Calgary4LIfe]

One thing about team CORSI that doesn't come up enough - or that is not as well explained as a reason to why it isn't THE stat that it is trotted out to be. Whenever there is an exception, the counter-argument is always: "That is what Toronto fans said too. That is what the Avs fans said too" regarding how teams allow perimeter shots.

Take the Flames. A young, less talented team than the contenders out there. No one is arguing against that (though I would argue they aren't quite as unskilled as most people keep stating in the media).

Watch their systems play in the defensive zone. They sit back a bit more. They are more patient in the defensive zone than a lot of other teams. It seems to me that Hartley has it drilled into them that once the opposing team has gained the zone, the team plays more of a positional defence. Sure, they do pressure the puck carriers a bit, but not as much as other teams.

Compare that to the Wild. They are very, very aggressive in the defensive zone. They don't give a team time or space out there. Without looking, I would just guess that the Wild have better CORSI stats as a team.

Consider Edmonton with their 'swarm' defence. Remember when they were losing and all those 'this guy scored' pics? Well, they really pressured the puck carrier, but didn't pick up any trailing men. They had great CORSI stats then if I remember, even though they were giving up golden opportunity after golden opportunity, and were rarely in a situation that someone could point and say: "They controlled the game, and they lost out of sheer bad luck."

I think where CORSI really becomes valuable is when you are comparing individuals on the same team. Same system (for the most part - the coach will obviously utilize guys a bit differently, and often there is a set play involved), same team, same season. You can see which players on your team further impact the team other than the regular stats page, or scoring chances. I think it is a wonderfully useful tool to compare players amongst themselves on the same team, and see where perhaps players should be facing lower quality of competition (sheltering) or players that perhaps you can trust against better competition.

I am just not sold on the team CORSI, as there seems to be too many exceptions. Calgary has been an exception all season long. Colorado was a huge exception last year - even winning the division. I am not convinced that CORSI was predictive in their collapse. Perhaps it was the roster changes that impacted the team more severely (losing Statsny, who I think was their best FO guy, and someone who seemed to be utilized like Backlund when they moved ROR to wing) hurt them more than helped. Maybe Varlamov is having a down year, and based on how relatively inexperienced their defensive corps are, that would have a big impact. Perhaps Roy has been trying to adjust how their team plays this year and tinkering with the systems. Perhaps the team is tuning him out a little bit?

Maybe it was predictive after all. I just don't see why it is some kind of a law. Leafs plainly sucked out there and were a poorly built team that was much like Calgary for the few years before the blow-up - constantly would go on many ups and downs, and most nights seemingly playing with half-effort.

As the OP pointed out, Anaheim has been a good exception as well to CORSI.

I can't say that the Flames won't regress next year. If they do, it is not because CORSI predicted it. I would bet it would be due to changes on the team somewhere. Maybe Treliving will demand a tighter defensive system much like Sutter did for Keenan. Maybe the roster moves that they will make will change the chemistry too much. Maybe Hartley will have players start tuning him out. Who knows... but it just seems like in a 30 team league, the amount of outliers is too great to have anything other than a casual correlation at best

Generally, yes, there is somewhat of a correlation with taking shots on net and winning a game. You can't win games if you don't take shots on net. What it continues to ignore is scoring chances (subjective) and quality of shots (again, shots from the slot = shots from the red line). The confidence level should decrease significantly, no?

CORSI (in my opinion) should be broken down into categories. Shots from the slot. Shots from the point. Shots of greater than 60 degrees. 2nd shots. 3rd shots. Shots through traffic. Etc... Sure, this is all way more difficult to calculate, but it would really be much more predictive.

But then again, a team will switch its' roster in the off-season, change up its' systems, and you would be left with a fairly different team on that stats paper. Even goalie save percentages can differ so much. Just look at Kipper's save percentages over the years. Look at Dubynk. Also, players themselves regress or improve. Players play through injuries, players become healthy, players are sometimes overly-favored by the coaches (and this is where advanced stats should really help to guide coaches here) or unfairly disliked by the coaches.

I like stats in sports. I believe they absolutely have a place moving forward, and it is yet another thing to look up and compare between players and teams. I just think they are in their infancy. It also seems like a lot of teams don't put a whole lot of stock into what we as fans are following in terms of statistical information, and have their own 'secret' stats that they utilize that are not only much more useful, but probably much more predictive.

Some stats guy wrote a nice post in one of the Flame-bashing threads on how the confidence interval was so low for CORSI, that he would be fired if he submitted work like that to his boss. He was a statistician of some sort. I was more or less sold on these current advanced stats right up until his post where he just ripped them apart. I should have saved it. If there are any statisticians on this forum (more than just guys like me who have had to take a couple of stats courses as part of their program, but guys who have majored in actuarial science), I am sure they could come up with an even better rebuttal to all things CORSI, PDO, and perhaps a few other ones that are being utilized. It is way beyond me.

[Geeoff]

Two big differences between us and Colorado are that they lost their #2 centre in free agency and Nathan MacKinnon had a sophomore slump, whereas Sean Monahan is having his sophomore year now and is killing it.

[DJones]

I've been thinking about it and one reason why advanced stats work so much better in baseball is because they base everything around getting on base. Getting on base is a tangible advancement for the team. If you get 4 of them in a row you get a point. That's why basing advanced stats on on base percentage can be so effective.

With Hockey they for some reason have chosen shot attempts to base everything around. Not goals, not scoring chances, not actual shots, god damn shot attempts. There is no tangible benefit for a team from each shot attempt. There is just the hope that the next one will count.

Until they solve that problem they will never be able to know the full scenario

[pylon]

Originally Posted by Quincy Egg

This post just reeks of ignorance. Ever wonder why every single NHL team employs some form of advanced stats in player evaluation?

This isn't the mid 80s.

It worked really well for the Oilers didn't it?

And secondly, just cheering for my home team as a fan, without having to look up a bunch of mathematical formulas and a scientific calculator on my coffee table is a lot more fun. If you are into the numbers of sports.... great. But I spend 9 hours a day working with numbers, the last thing I want to do is to suck the joy out of a Hockey game, by going back to work.

[Street Pharmacist]

Originally Posted by pylon

It worked really well for the Oilers didn't it?

It can't possibly have helped in one season as it's for player procurement and drafting. That's a pretty lousy example.

Treliving has stated they use their own Metrics similar to corsi, etc. Did it work for the flames?

[Robbob]

Now that you mentioned it. Both teams that were public hiring these web number crunchers sure are doing well.

I think it goes without saying that all hockey teams have advanced stats that they don't disclose. It really is the logical next step in an industry where most things are on a level playing field.

I think the Oilers are showing that you can't build a team on players with high CORSI numbers as there is a lot more to it then that. If these fancy stats were the end all then you would have guessed that the oilers should have taken a step forward this year after signing wonder CORSI players like Fayne and the other guys (forget his name).

[Street Pharmacist]

Originally Posted by Robbob

Now that you mentioned it. Both teams that were public hiring these web number crunchers sure are doing well.

I think it goes without saying that all hockey teams have advanced stats that they don't disclose. It really is the logical next step in an industry where most things are on a level playing field.

I think the Oilers are showing that you can't build a team on players with high CORSI numbers as there is a lot more to it then that. If these fancy stats were the end all then you would have guessed that the oilers should have taken a step forward this year after signing wonder CORSI players like Fayne and the other guys (forget his name).

The only good corsi hit brought in really was Fayne. Nikitin was terrible and the rest of the Oilers were already the worst on the league. The oilers don't show anything.

Not that I believe corsi is the new all and end all anyways, but again, poor example

[Robbob]

Pouliot is who I was thinking of. From a puck Daddy article headline:

"Oilers sign advanced stats darlings Benoit Pouliot, Mark Fayne"

[Street Pharmacist]

Originally Posted by Robbob

Pouliot is who I was thinking of. From a puck Daddy article headline:

"Oilers sign advanced stats darlings Benoit Pouliot, Mark Fayne"

My point is they were terrible before, and bringing in two new "advanced stats darlings" and failing is hardly a proof of anything other than Edmonton is No Good

[Robbob]

That is the thing though. For the last few years we have been hearing that the Oilers have good fancy stats and that they are just unlucky. You would think if that was the case then adding a few more fancy stat players would improve their team rather then turning in one of their worst years.