twabby
Registered User
- Mar 9, 2010
- 13,772
- 14,714
I don't think it's as significant a difference as you do. Was he extremely lucky earlier in the year? Does that affect your calculations and expectations?
And as I said, PDO is ASSUMED to measure luck. It's not actually a measure of luck. It could easily measure "miscellaneous factors" that could include a player's actions, changes in behavior, or just about anything.
Even your corisica egf formula is incomplete and suspect for various reasons, not the least of which is the exclusion of blocked shots and the assumption that shots from closer range increase goal chances in an improving, non-linear manner. In REAL LIFE, not Stat-land, proximity to the goaltender also cuts down on your angle if the goaltender is in position for the shot. So a 10 foot shot may actually have less net open than a 15 foot shot depending on where the goaltender is. How far is the goaltender out of position laterally on the shot? Is that accounted for, and wouldn't just one or two goals like that make a big difference over 20 games? Are all rebounds weighted equally, even though some never have a chance of going in while others are tap ins? Once again, we're using a functionally oversimiplified (however algorithmically complex) shot-based stat manipulation to try and explain a complex game full of unquantifiable factors.
To try and project expected goals for 20 games based on that formula, and cite that as proof that PDO is correct, is folly. You use his career average to prove this, which is also a small sample size that includes his "hot" period earlier in the year as a heavily weighted portion of the available data.
Overall your logic is going like this:
It's luck
How do you know?
Because PDO says so
How can you prove it's really luck?
Because PDO measures luck
That's circular. Especially since you're assuming the differences in stats that comprise a PDO calculation are explained by luck, but you say "specific evidence to suggest he only had poor quality shots" are required to knock you off your bad luck stance. So you assume one intangible thing (luck via PDO) but reject another immeasurable (that you at least try to account for, however imperfectly, and only when it backs your argument).
So change the word from "luck" to "variance" if you want to account for miscellaneous factors, it doesn't change the argument. The fact is PDO is highly variable and inconsistent across pretty much every NHLer and except for very extreme cases, most players have PDOs that fall between 99 and 101 over their careers. Prolonged stretches of PDOs outside of these numbers simply don't last. I'm willing to guess that Kuznetsov's will not remain at 96, despite txpd saying maybe he's just a hard-luck player.
And while using his career average, it also included his cold streak. I didn't selectively decide to ignore his cold streak when looking at his career shooting percentage, similar to how it would be silly to ignore a goalie's cold streak when looking at his save percentage.
At this point I have no faith in your ability to even approach statistics from a rational standpoint. You're talking about the expected goal formula like it was just plucked out of thin air instead of constructed based on past data (which it was). No, it's not perfect. Nothing ever will be. But just because a formula/measure is incomplete doesn't mean it's not useful.
It's an interesting double standard because while the eye-test cannot fail (despite it being subject to huge amounts of bias and other factors such as not being able to feasibly watch every single game of every single player in the NHL), if a statistical model is created and it doesn't 100% account for every small factor then it is completely null and void in your eyes.