Friedman's 30 thoughts - analytics say Oilers' luck should turn

[Resolute 14]

Originally Posted by nik-

When all you shoot is muffins from the perimeter and you give up five 2 on 1's a game, your PDO is going to suck.

Sorry, that's not luck.

And, as always, bias colours conclusion. Edmonton's shooting percentage at 5 on 5 is 6.60%. Ahead of only Florida, Columbus and Winnipeg (who are at an atrocious 5.65%). If one continues to buy into the "Edmonton has so much talent on offence" argument, then one would interpret that shooting percentage as being bad luck.

[liamenator]

I have a couple of elementary but earnest questions about fancy stat methodologies that I am sure are addressed multiple places, but I don't have the motivation to wade through the literature right now.

How are shot attempt data collected, i.e. who collects them? Are shot attempt metrics like Corsi reliant on official NHL statisticians? If so, how do they account for variables like the well-documented differences in how official stats are compiled in various rinks around the NHL? A shot in Columbus ≠ a shot in Calgary; a blocked shot in SJ ≠ one in New Jersey.

Or are these data tracked and compiled by analytics guys themselves? I get that each team probably does their own tracking. But where do the data used by bloggers come from?

[octothorp]

I think advanced stats for hockey are really in their infancy, compared to where other sports are at, and a lot of that is that hockey is one of the most difficult sports to statistically quantify, and right now, most of these models are incredibly simplistic, including PDO.

But that said, Freidman really screws up part of his explanation: luck does not even out, nor does PDO suggest that it should (although right from its inception, its been linked to the idea of luck evening out). PDO merely suggest that there will be a regression toward the mean, especially for teams that look like outliers. It certainly doesn't make the claim that a team that has been lucky early in the season will be unlucky later in the season, or vice versa. Only that luck, going forward, will be average throughout the league.

But here's the big problem with these scores: there's no reason to expect the order of teams to change, since regression affects all teams; it will simply affect outliers more. There's nothing there to suggest that the Flames won't finish with the second highest PDO at the end of the year. And there's nothing to suggest that the Oilers won't finish with the lowest PDO at the end of the year. You'd simply expect teams to be bunched more closely together at the end of the year.

If all of that is a big problem, then here's the massive, colossal problem with that expectation: the current distribution isn't that out-of-line with the year-end distribution, which typically has the top few teams in the mid 1.20s, and the bottom few teams in the .970s (with the occasional outlier way down in the .960s). So if we expect a similar cluster of scores this year, there's nothing to suggest that Calgary or Edmonton (or any other team) actually will regress toward the mean, since all fall within those expected distributions. Some will regress toward the mean, others will progress away from the mean, others will remain essentially unchanged.

Now, you can break down the two numbers, and argue that there's slightly more regression to the mean for shooting percentage than there is for save percentage. Traditionally, the top teams in the league have a shooting percentage of around 0.10, and both Calgary and Tampa Bay may be expected to trend downward slightly over the course of the year. But they aren't really out of line with the top teams in 2012-13. So it would take a lot more analysis than I could put into it to suggest how likely it is that those numbers will hold and how much they will regress.

[edslunch]

I hate the concept that PDO is a measure of luck. Sure there is some, but a bad goalie will have a bad sv% luck or no luck, same with shooters and systems and execution. I get that all these players are at the extreme tail of the bell curve so are very similar, but still there are differences beyond luck

[monkeyman]

I think advanced stats has a place during the course of a regular played game in which players and teams aren't conscious of it. It really just tells you who's carrying the play and creating more chances, which is usually the better team.
I think the mistake is when you try to play to it as a way of making your team better. Instead of passing to someone in a higher percentage scoring position, you take a shot, instead of driving to the net, you take a shot... sure, you'll increase your Corsi and Fenwick but you're likely turning the puck over more and generating less quality chances. I think you'd likely make your team worse if you were consciously trying to play for positive Corsi and Fenwick.
So I guess I’m saying there are a natural Corsi and a contrived Corsi. I think if teams are moving toward the idea that taking more shots will automatically make them better(contrived), Corsi will become less an indicator of how good or bad a team is.

[edslunch]

Originally Posted by djsFlames

That these stats can argue FOR the Oilers (claiming 'bad luck') and against the Flames, says enough about how much weight they should carry.



Yes, things average out. And the Flames unsustainably high percentage goaltending did fall back to earth, but it did not affect the consistency of them winning games.



I know our shooting percentage must be well above average, but if you look at where the majority of our shots are being taken game after game, versus our opposition, we're often shooting from much higher probability locations on the ice. We're averaging less shots per game than most, but making our shots count because the majority of them aren't from the perimeter.

Are shot positions available online as data, not just charts? It ought to be easy enough to come with some sort of weighted shot quality star based on distance and angle to the net

[Jason14h]

It is a similar stat to baseballs BABIP (Batting average on balls in play)

When a player has a much higher average on balls in play then their historical average it is seen as an ourlier.

HOWEVER: Good players who hit the ball harder have a better BABIP. This is the issue with PDO.

On HFboards it showed the Oilers over the past 6 years. 4 years they were bottom 5 in PDO and never cracking top 10.

So the Oilers must always be unlucky!

...........OR they give up above average scoring changes and generate below average scoring chances.

Where the Flames may regress would be if teams learn to defend the fast break from the defense. The Flames generate very few goals that are not off the rush. Conversely, they give up very few odd man rushes which historically result in good scoring chances.

[Resolute 14]

Originally Posted by liamenator

I have a couple of elementary but earnest questions about fancy stat methodologies that I am sure are addressed multiple places, but I don't have the motivation to wade through the literature right now.

How are shot attempt data collected, i.e. who collects them? Are shot attempt metrics like Corsi reliant on official NHL statisticians? If so, how do they account for variables like the well-documented differences in how official stats are compiled in various rinks around the NHL? A shot in Columbus ≠ a shot in Calgary; a blocked shot in SJ ≠ one in New Jersey.

Or are these data tracked and compiled by analytics guys themselves? I get that each team probably does their own tracking. But where do the data used by bloggers come from?

I believe that largely, they use the NHL's stats. And yes, like shots and hits, that makes them susceptible to the same biases of the home town scorer. But the belief is that over time - as the sample sizes get larger - those individual outliers are absorbed into the whole and their impact on the overall numbers become insignificant.

[AcGold]

The Oilers did a line change when there was a breakaway on Scrivens, think about that. You don't get good luck when you play like morons.

[CaptainCrunch]

We're so unlucky

Nobody loves us

We can't catch a break

My kneecaps hurt.

[Locke]

The only way they'll get enough 'luck' to turn that team around is if they have horseshoes shoved so far up their asses they start coming out of their eyes while eating four-leaf clover salad.

[FlameZilla]

Originally Posted by _Q_

I'm convinced that Hartley's system is his way of trolling the analytics folks.

I hope he continues with this for 5 years, wins a couple of Stanley Cups and makes a bunch of heads explode in the process.

"Advanced stats" are no good.

Hartley's system is the classic "counter-attack" method. Possession stats don't mean much when your team is built to catch other teams on the break with speed. Depth, youth and conditioning are other big advantages our team has this season.

[trublmaker]

Luck is about all they have left to hang onto. No good

[Resolute 14]

Originally Posted by Jason14h

On HFboards it showed the Oilers over the past 6 years. 4 years they were bottom 5 in PDO and never cracking top 10.

So the Oilers must always be unlucky!

...........OR they give up above average scoring changes and generate below average scoring chances.

The Oilers are undoubtedly terrible, but looking at each season over the last bit does give an indication of why people look at the numbers for Calgary and Edmonton and contemplate luck. At 5 on 5 according to hockeyanalysis.com:

             Calgary     Edmonton
2014-15  102.4 ( 3rd)    96.6 (30th)
2013-14   98.8 (26th)    99.0 (24th)
2012-13   97.2 (29th)   100.6 (12th)
2011-12  100.2 (12th)   100.1 (13th)
2010-11  100.0 (18th)    98.7 (28th)

Not only are both of our teams huge outliers when compared against the other teams this season, but we are huge outliers compared to our own performances in recent seasons. It is not hard to blame people for thinking both teams will trend toward the middle at some point.

[dash_pinched]

I always believed that luck is the residue of hard work.

[Calgary4LIfe]

Seeing Calgary as a terrible CORSI and above-average PDO team, and then reading about how they will regress was insightful.

WAS.

Seeing the Oilers be above average in CORSI and while having the worst PDO and having these experts claim the Oilers are going to be better has changed my mind.

There is a reason why teams like Chicago have developed and use their own in-house metrics. I really used to believe in CORSI, though I honestly wasn't even close to a fancy stats guru by any means. But seeing two contrasting teams in the NHL for this long now, on opposite ends of that scale? Screw those metrics. My eyes tell me the Flames play a very good game all-around. They are great on the rush, they are great in the neutral zone, and they are great in the defensive zone. Goalies have stopped playing out of their minds after about game 5 IIRC. Flames win by out-working and out-chancing their opponents.

Oilers fail miserably in every zone - yes, even in the offensive zone. Their PDO sucks because THEY DO SUCK. Their CORSI is great due to score-effects. They are continually out-chanced and dominated regardless of what their CORSI says happened. What you see and feel in an Oilers game is NOT congruent to what CORSI says.

I bet teams in the NHL do look at CORSI, but I also bet that whatever fancy stats they use in-house is so much better as a whole. I bet their numbers say that the Oilers are really crappy, and I bet they also show that the Flames are perhaps an above average team at least. Those are what my eyes say anyways.

[nik-]

well, to be fair, betting on the Oilers to get better is pretty easy. They legitimately can't get any worse, so there's only one way to go.

[Scornfire]

Originally Posted by nik-

well, to be fair, betting on the Oilers to get better is pretty easy. They legitimately can't get any worse, so there's only one way to go.

How many years have we said that, and yet...

They keep it up and they'll be resorting to open tryouts, just imagine it, a team full of local players wearing sweatpants under their equipment and facewashing the other team with Dorito dust

[ColoradoFlamesFan]

Dear Mr. Lowe and Mr. MacTavish:
Please stay the course, your "luck" will change. Hall will start to back check, and Shultz will stop having brain cramps, no really....
it'll happen, stay the course
your Friend CFF

[tvp2003]

Edmonton: Even our luck is no good.