If someone's shooting percentage is 37% I will predict that it will go down.
That doesn't make it a predictive stat. It just means that a guy is way outside the normal range so its a no brainer to predict regression. If that does make it a predictive stat in your mind, then every stat is a predictive stat.
Without any other context than simply looking at the figure relative to some average, sure - it will probably go down.
With context? Maybe it will still go down, back to the "average" figure. [Which pretends that the figure we cite is the true, correct mean going forward - but that's a discussion I'll save for another time.] Maybe the guy spent a shitload of time this offseason working on shooting accuracy and pinpointing corners, and he's been able to apply it - so, it might still go down but not as much as one might think. Or maybe the guy is sitting in a high-percentage area that suits his style of play even more, and for him 37% is his new norm. Maybe he's even underperforming what he really should be doing in that spot. Without context for why the shooting percentage is 37% and what we should really expect on average and why a player might be better/worse than average, no one knows whether it's truly sustainable or not and what we should expect going forward.
That's really the problem with every "regression to the mean" claim: it pretends to know what "the mean" is at any and all given times, assumes that
all of the known information is baked into a couple numbers that can be compared, and no further context need be given to those numbers. And frequently, it only assumes "regression to the mean" and not an overshoot/undershoot in future performance which might bring about yet another "regression to the mean" claim but now in the opposite direction.
Bet his stats didnt predict the 3 spankings the Canucks gave Oilers early this year. This is why i love hockey lol. Luck is an important part of the game as well but stat nerds keep ignoring those factors.
I prefer "randomness" to "luck" but I'll completely admit, that's nitpicking by invoking a statistical concept in place of some fate-based concept.
This is not wrong, but I feel like it's a bit backwards. People familiar with statistical concepts and probability distributions say "the bottom is true, but the top is what's most likely to happen." People who are bad at statistics see the bottom and say "see, anything can happen" and then when a low-probability scenario occurs, it's "proof" that "you so-called experts know nothing about statistics" and then becomes an argument for why that last outcome is not just highly predictive of what will happen next, it's "proof" that the same event will happen next; the further the observed event falls into the tail of the probability distribution, the more everyone should believe it will definitely happen again.
www.thesportster.com
