Quote:
|
Originally Posted by JPSeraph
How does that which is in the "realm of opinion" (i.e. defense) factor into a player analysis which is based primarily on quantitative measures?
Suppose, for instance, that we agree to a 50/50 split between offense and defense with respect to how each category should factor into a player's overall value. If the measures for 50% of a player's value are, as you say, merely opinions, then how should they be weighted against the other 50% which is more quantitatively sound?
If they are weighed equally, it seems contradictory to your stance that quantitative measures should be the primary (or is it only?) factor in such an analysis.
|
That isn't my stance. My stance is to accept numbers over observation when there are measures available that seem to properly quantify the issue in question, removing most or all confounding factors.
There are such measures for scoring, rebounding, passing (though far from unquestionable), so I prefer them over using observation to decide who was better. Defense has no such numbers (though there are some interesting ones), and therefore, there's a lot of subjectivity involved.
But defense is clearly a significant part of the game. That's the difference between it and intangibles. We know defense is a huge part of the game and a big determining factor in who wins and loses. We don't know, at all, that intangibles are a huge part of the game and a big determining factor in who wins. Some people may theorize that it is, but with no proof. It's akin to arguing whether God exists: there's no way to prove it one way or the other.
Defense exists, though. I don't think anyone denies that.
Quote:
|
But is that what Hollinger did?
|
No. Hollinger, I believe, builds measures by what he considers logical and then evaluates them by standards like repeatability: do players tend to repeat their PER performance from season to season? Doing so is evidence that it's getting at their talent level, rather than being somewhat random, because players repeat their talent level from year to year, so their PER should (by and large, with accepted statistical variance) by similar from year to year (adjusted by age).
That's far from unquestionable, as I said above. But I think his logical assumptions (like that scoring and assists are affected by team pace and rebounds are affected by how many rebounding chances there are, etc) are strong ones.
Quote:
|
I know for me personally that the joy of statistics is watching the numbers describe what I see in reality, or vice versa.
|
Same same.