Against PDO

[Just Linda]

Could you please elaborate what you mean by "micro" and "macro" (in this context), I'm lost?

Micro - specific situations for specific players
Macro - broad sweeping generalization

 

[MadLuke]

Exactly! One does not need PDO to see that 14-1-1 streak is ... a streak, it can't be maintained over the whole season. Why then rely on PDO?

Not sure this is a good way to look at it, if I understand the corsi/PDO type of claim, it is if they would achieved that streak by outshooting their opponent 2 to 1 it is more likely that they sustain a larger sample size to that rate that if they did it with while being outshoot 0.98:1 with an very high shooting percentage and save percentage. If they would did it with a 99.5 PDO, maybe it could be maintained.

I tend to agree that assuming a team with great goal tending should necessarily regress in save percentage sound flawed.

Has for the claim that PDO do not tend to trend toward 1 has sample size goes up:

If we look at last year around this time of the year (November 30, teams played in average 25 games):
https://web.archive.org/web/20181130071011/https://www.hockey-reference.com/leagues/NHL_2019.html
vs
And at the end of the season
2018-19 NHL Summary | Hockey-Reference.com

Top 5:
Maple leaf: 103.1 vs 101.6
Islanders: 102.5 vs 101.9
Avalanche: 102.4 vs 99.9
Capitals: 102.1 vs 101.8
Predators: 101.7 vs 100.4

Bottom 5:
Sharks: 97 vs 98.9
Knights: 97.7 vs 99.4
Canes: 97.9 vs 99.4
Coyotes: 98.0 vs 99
Panthers: 98.1 vs 98.7

All of the top 5 team's in PDO by the end of November trended down toward 1.0 for the rest of the year, the reverse for the bottom 5.

Would need to maybe do it with different time range and for sure many season, but to expect the top and bottom PDO to get back toward has more game get played seem a solid assumption.

 

[BrockLobster]

Its weird how its always the best teams with high PDO and the worst teams with the worst. Its almost as if PDO is tied to skill!

 

[Garbage Goal]

I don’t get what the post is meant to achieve. It’s constructed on strawman and begins with an extremely incongruent comparison. On top of that the OP seems to either believe everyone believes these myths or he’s attempting to address a sub-section of people who miss the point of using analytics as a tool entirely. Except he’s talking rather condescendingly and preachy so that target group, which isn’t likely to listen in the first place, probably isn’t listening.

Anybody worth talking to already knows to use every advanced statistic as a tool and knows there’s a pretty normal range outside of 100 for PDO that shouldn’t be ascribed purely to luck.

The good old gamblers fallacy. Clearly if I flip a coin after getting three tails, there is less than 50% chance that next coin is Tails, so bet Heads, baby.

Or as I like to call it... "The HF regression to the mean" fallacy.

That has less to do with the gamblers fallacy. That’s just combinations versus permutations in statistics.

Any individual coin flip has a 50% chance, but you only have a .05% chance of flipping 11 tails in a row, for example. So, yes, if the three heads hadn’t already been flipped the smart bet would be you won’t flip four heads. Since the first three already hit heads it’s a 50/50 chance at that point. The chances of hitting four heads in a row are still significantly less than 50% though.

 

[7even]

So, let's give a rest to PDO as a measure of luck, leave it to uneducated folks like Lambert and Luszczyszyn.

 

[Hunn]

Not sure this is a good way to look at it, if I understand the corsi/PDO type of claim, it is if they would achieved that streak by outshooting their opponent 2 to 1 it is more likely that they sustain a larger sample size to that rate that if they did it with while being outshoot 0.98:1 with an very high shooting percentage and save percentage. If they would did it with a 99.5 PDO, maybe it could be maintained.

I tend to agree that assuming a team with great goal tending should necessarily regress in save percentage sound flawed.

Has for the claim that PDO do not tend to trend toward 1 has sample size goes up:

If we look at last year around this time of the year (November 30, teams played in average 25 games):
https://web.archive.org/web/20181130071011/https://www.hockey-reference.com/leagues/NHL_2019.html
vs
And at the end of the season
2018-19 NHL Summary | Hockey-Reference.com

Top 5:
Maple leaf: 103.1 vs 101.6
Islanders: 102.5 vs 101.9
Avalanche: 102.4 vs 99.9
Capitals: 102.1 vs 101.8
Predators: 101.7 vs 100.4

Bottom 5:
Sharks: 97 vs 98.9
Knights: 97.7 vs 99.4
Canes: 97.9 vs 99.4
Coyotes: 98.0 vs 99
Panthers: 98.1 vs 98.7

All of the top 5 team's in PDO by the end of November trended down toward 1.0 for the rest of the year, the reverse for the bottom 5.

Would need to maybe do it with different time range and for sure many season, but to expect the top and bottom PDO to get back toward has more game get played seem a solid assumption.

What these data show is that PDO distribution is narrower in April than in November. What they don't show is that PDO converges to 1 (because it does not).

 

[93LEAFS]

The flaw in this premise is, that a player who plays 82 games against NHL goalies is going to have his oiSH% generally fall within certain thresholds that can be deemed reasonable. Yes, elite players are going to drive up shooting percentages, but not over certain marks. I assume one player playing behind an absolutely terrible goalie for 60 games could kill his oiSH%, but even then, its hard to see it consistently fall below 87% of so. The first mistake is looking at it in absolute terms, instead of with nuanced belief. I mean, a Mitch Marer/JT line or a Nylander/Matthews line is going to likely shoot a higher percentage than a Nick Shore/Gauthier line. That doesn't mean it would be reasonable to expect one to break historic league norms and the other to perform dramatically below league norms.

 

[ricky0034]

*cough* 2014avs *cough*

sorry must be feeling a little sick

 

[Hunn]

*cough* 2014avs *cough*

sorry must be feeling a little sick

2014 Avs PDO was within "typical" 98-102 range. It was not even the highest PDO in that season, but people tend to forget that because it does not fit into the narrative discussed under "Myth No 3" section.

 

[MadLuke]

What these data show is that PDO distribution is narrower in April than in November. What they don't show is that PDO converges to 1 (because it does not).

Get narrower around a value and converging toward a value do sound like 2 ways of saying the exact same thing.

 

[Hunn]

I don’t get what the post is meant to achieve. It’s constructed on strawman and begins with an extremely incongruent comparison. On top of that the OP seems to either believe everyone believes these myths or he’s attempting to address a sub-section of people who miss the point of using analytics as a tool entirely. Except he’s talking rather condescendingly and preachy so that target group, which isn’t likely to listen in the first place, probably isn’t listening.

Anybody worth talking to already knows to use every advanced statistic as a tool and knows there’s a pretty normal range outside of 100 for PDO that shouldn’t be ascribed purely to luck.

I do think that considerable fraction of hockey fandom believes in 3 myths outlined in OP. I maybe wrong, probably it's just a couple of bloggers. As about bolded, fair point, I apologize for that.

 

[Hunn]

Get narrower around a value and converging toward a value do sound like 2 ways of saying the exact same thing.

It sounds indeed like that, but it's not. Wider distributions at the beginning of sampling process converge to final distribution that may or may not be narrower, but it's not required to be delta-like (i. e. non-zero at one value only).

 

[hotpaws]

just another useless stat

 

[Avelanche]

PDO is a stat you glance at and it tells you to look deeper into a team/player. If you can’t explain it, most likely they are unlucky/lucky.

 

[McDNicks17]

PDO is weird but OP's argument is also weird.

This is more or less what I was thinking of saying:


To reiterate: Why in the world would you add together two unlike things? This is like saying 3 kilometers + 4 grams = 7 squirrels.

If you have a goalie who's letting everything in all season (as a Red Wings fan I would know a lot about this), PDO goes down. What does that have to do with anything? Why would you make that a team stat? Why would you relate it to shooting %?

If you have a player who's very good at picking his shots, PDO goes up for him and his linemates. Okay? What does that have to do with the team's save %?

If one player does a very good job of clearing the crease but has zero offense, and another player cherrypicks all game and snipes well, they both get the same PDO? That just seems really awkward and useless.

Because it was originally created to add context to +/- and other goal% based metrics.

 

[McDNicks17]

People need to realize PDO was one of the first non-traditional stats created. It came about back when +/- was still one of the main ways players were evaluated. It isn't some super complex stat created to be the be-all-end-all of predictors. It's simply a quick reference point for context when looking at goal differentials.

It was basically made for when people would say stuff like "Mark Fistric was +27. He's really good!". You could then take a quick glance at his PDO and assume he would never do that again.

 

[Stamkos4life]

Could you please elaborate what you mean by "micro" and "macro" (in this context), I'm lost?

Dont worry, his post made no sense.

Like Alberta Einstein said, any fool can make something more complicated than it needs to be.

 

[TheOtherOne]

People need to realize PDO was one of the first non-traditional stats created. It came about back when +/- was still one of the main ways players were evaluated. It isn't some super complex stat created to be the be-all-end-all of predictors. It's simply a quick reference point for context when looking at goal differentials.

It was basically made for when people would say stuff like "Mark Fistric was +27. He's really good!". You could then take a quick glance at his PDO and assume he would never do that again.

But it doesn't even work in that context because you can draw opposite conclusions from the same data.
- The PDO is high due to luck and is therefore unsustainable.
- The PDO is high due to the fact that he has excellent shot accuracy so as long as his shot doesn't fall off he will sustain it.
- The PDO is high because he plays in front of a HoF goalie or an excellent shot suppression system so it will sustain as long as he plays on the same team.

What is the use of a stat that supports opposite conclusions?

 

[McDNicks17]

But it doesn't even work in that context because you can draw opposite conclusions from the same data.
- The PDO is high due to luck and is therefore unsustainable.
- The PDO is high due to the fact that he has excellent shot accuracy so as long as his shot doesn't fall off he will sustain it.
- The PDO is high because he plays in front of a HoF goalie or an excellent shot suppression system so it will sustain as long as he plays on the same team.

What is the use of a stat that supports opposite conclusions?

Extreme examples existing doesn't mean the stat "doesn't work".

 

[Video Coach]

1. All NHL players are roughly equal in skill, to the point that there are not vast skill disparities between different clubs or players. Thus:
2. All shots and shot attempts are equal. Thus your shooting percentage is a function of luck.
3. All saves and save attempts are created equally. This your save percentage is a function of luck.

This.

I always hate when I hear the "he's shooting too high a %, he's obviously going to regress." Sometimes you're shooting the puck really well. And generally if you're shooting well, you keep it up because you feel confident, so your shooting is more sure. The opposite is true as well, if you're not scoring, you're likely going to be less confident and not shoot as well.

Some guys have better shots than others. Victor Mete didn't score for 2+ seasons because his finishing ability wasn't good enough. He'd have an open chance on a 2 on 1 and slide into the goalie's pads instead of burying it in the top half of the net. That's not bad luck, that's poor shooting. But he worked on his shot extensively this offseason and he's got 3 this year.

To me, shooting % is largely determined by 3 factors:
- Where the shot is coming from
- How good the shooter's shot is
- How hot the shooter is

Goaltending and "luck" is somewhat of a factor, but not much. Usually the goals that go in are ones that should go in - ie great shot, open net, deflection, rebound/scramble. I watch a lot of hockey and don't see a ton of "lucky" goals. Sometimes a goalie lets in a stinker, or a puck takes a crazy bounce...but not that often. Most pucks get stopped because the shot was from a low % spot on the ice, and/or the shot wasn't good enough.

 

[TheOtherOne]

Extreme examples existing doesn't mean the stat "doesn't work".

What is extreme? The existence of luck? Or the existence of players who shoot better than other players?

 

[McDNicks17]

What is extreme? The existence of luck? Or the existence of players who shoot better than other players?

Only elite players have the ability to noticeably influence onSH% singlehandedly and there's obviously very few players playing in front of HoF goalies.

Both your examples are extreme cases.

 

[x Tame Impala]

Seems like an attempt to quantify a small sample size event to larger quantities of data. As well as what @tarheelhockey said, I don’t see what benefit it offers that looking at a player’s current SH% or point production is and comparing it to their career.

If player X is struggling and shooting well below his career average shooting percentage it’s perfectly reasonable to assume they’re going to play better at some point and get closer to their career averages.

I also don’t see how their could possibly be such a substantial correlation between SH% and SV% that the two can be combined into one stat. I’d actually really appreciate if someone can make an argument for that because I think it’s pretty silly at this point.

 

[tarheelhockey]

Disagree.

1) PDO is meant to show how sustainable a teams play is. No different than when a player goes on a hot streak and shoots 40%, you know eventually it will even out. Look no further than Buffalo, they were a high PDO team through the first 10 games and then regressed back to 100 over the past 12.

So look at their shooting% and decide whether it’s sustainable or not. That’s pretty a simple way to judge that question.

Adding their shooting% to their save% and measuring how far the combined total is from an arbitrarily-selected 100 is... I mean, that’s just silly. There’s no sound basis for judging sustainability that way.

2). Sure there is. A shot from the side of the net from along the boards that has almost 0 chance to go in shouldn't be counted the same as someone right in the slot taking a shot.

I would argue PDO is better than it ever has been with parity what it is. In general, teams are going to score generally the same and have goaltending generally the same. Over an 82 game season these things usually even out, but where PDO shines is even bigger sample sizes than that.

In fact, using PDO is almost always a regression litmus test for teams who unexpectedly shot up the standings.

And again, you can better make these predictions by simply using the basic stats that were combined to create PDO.

A team shot up the standings with .950 goaltending? Obviously they are due for regression when the goalie cools off. There’s no need to add their shooting % and measure from 100 to draw that conclusion.

 

[Kamiccolo]

So look at their shooting% and decide whether it’s sustainable or not. That’s pretty a simple way to judge that question.

Adding their shooting% to their save% and measuring how far the combined total is from an arbitrarily-selected 100 is... I mean, that’s just silly. There’s no sound basis for judging sustainability that way.

Hmm, are you talking about indivdual PDO or team PDO? I am talking about team PDO, which encapsulates the team shooting % and save %.

And again, you can better make these predictions by simply using the basic stats that were combined to create PDO.

A team shot up the standings with .950 goaltending? Obviously they are due for regression when the goalie cools off. There’s no need to add their shooting % and measure from 100 to draw that conclusion.

Sometimes it isn't that simple though. Your argument is basically PDO sucks because if you wanted to see if something was sustainable, you would just check the sv% and the shooting% of the teams. But that is exactly what PDO is.