fuzzy_dunlop wrote:^
a) LOL didn't we have the exact same argument about steals, with you being vehemently on the other side of the aisle? But fine, it doesn't measure spacing per se but it helps spacers. Korver is 18th in OBPM, that doesn't seem far off 2 me.
b) I desperately want PI because NPI is awful. It's not a very subtle argument but I think it overwhelms any more nuanced second order effects.
c) Those figures came from a throwaway JE tweet that didn't include offense/deefnse splits... and I WAS talking about VORP in the general sense. AD can only be replaced by a big 99.99% of the time, so it makes perfect sense to "adjust" his offensive impact thus (and of course the same adjustment should then be made defensively).
a) No I was on the exact same side of the argument: That the box score stat (steals, 3PG, whatever) is a flawed measure of the actual impact a guy has when working on that front, oftentimes to a huge degree. The fact that it means I say one guy is being underrated and another is being overrated doesn't remotely mean I'm contradicting myself.
Re: Korver's OBPM doesn't seem far off to you. Right so you're choosing the method that agrees with your hunch, and therefore coming to the conclusion that that method must not be underrating the guy...which is another way of saying you aren't using any metric at all really, you're just coming in with an opinion, and grabbing on to stats that say you're right.
I'm not actually alleging that's all you do, but that's the danger.
b) Because NPI is awful, and PI is better. Based on what? Your opinion of how players actually should be? That's not good enough.
Look, the way to use these stats is to understand where there is potential for issues, and to adjust your confidence levels accordingly. If you see a set of circumstances in which PI would give you much more confidence than NPI, then yes, you should really want to see the PI, but to simply "wait for the PI" whenever you see NPI values you don't like is quite problematic.
To put it another way: I don't use PI instead of NPI, I use them both. Typically the PI is the metric I have more confidence in, but not always. To know the full story though, I won't both, plus more raw +/-, plus RPM, plus more pure box score metrics. That's not a feel-good "Everything has something to offer!" statement, it's simply a truth that to max out my confidence, any particular metric or tally I see makes me ask a question that another metric or tally can help me answer.
c) "Those figures"? Hmm. So you weren't referring to Evan Z's complete set of NPI, you were referring to JE's lone tweet. Okay. I get your confusion now, but doesn't really change the reasons why I posted the numbers I did though.
Re: "Davis can only be replaced by a big so it does make sense to adjust." No, you're not understanding. To oversimplify for sake of clarity:
If there were a 2-on-2 NBA league, and one team had 2 pairs of players: 2 guards who substituted for each other, and 2 bigs that substituted for each other. And we saw the following measures:
Guard 1 +7
Big 1 +4
Big 2 -4
Guard 2 -7
"adjusting" as we're talking here would mean giving a +3 boost not simply to Big 1, but to Big 2. And if you're doing that, by what basis would you adjust all Bigs?
None. Look when folks say stuff like "Point guards aren't DPOY level defenders", what they are actually saying is "The amount of drop off you get when you substitute one point guard for another on defense just isn't that big." It makes no sense therefore to "adjust" your measure and make it so that the #1 point guard defender ranks just as highly as the #1 center defender, because one guy can transform his team with defense, and the other guy cannot.
Were we to extend the discussion to another sport, what you'd be saying is essentially:
"If a measure of impact rates the top quarterbacks ahead of the top running backs, therefore we need to adjust the running backs metrics up so that the #1 running back is seen as just as valuable as the quarterback." And such would be silly. There's no reason to assume that all roles are as tough to replace as others, which is why no one does. The salaries for the different positions in football vary greatly, because it's well established one shouldn't adjust like this.