NET moves to 91
- Thread starter bac2therac
- Start date
It’s not really “one” point though. From an efficiency standpoint it’s 67 possessions out of however many there have been this year so far (~1300?). There’s just no compelling reason those 67 should be thrown out.
With all due respect (which is quite a lot given your background), this is absurd.
The people who risk money on this stuff have actually checked, and efficiency models and other models that incorporate margin of victory are vastly superior in terms of prediction quality to models that ignore that information. I understand it’s not “clean” from an academic standpoint (yes, the point of the game is to win, not rack up points). I am sympathetic to arguments that this should not count for postseason selection. But from a pure predictive modeling standpoint, ignoring victory margin is completely idiotic.
Lol, fluoxetine.
OK, let me be clear. I was hyperbolic a little bit. But you highlight my point.
Predictive models ought to be distinct from models that describe cumulative success.
Efficiency, margins, etc are critical for predictive models, along with stages of the moon, start time, number of days off, etc depending upon the team.
Describing how well a team has done, has to be bounded tightly around wins and losses and who you have beat and to a lesser degree, who you have lost to.
That latter is what I was alluding to when I said binary.
Cheers,
Loyal
Ok, yes, I think I mostly agree with you then. There are sort of two conversations going on in the thread which are getting crossed.
A lot of people seem to think the NET is not correctly measuring what it is intended to measure (which seems to me to be mostly net efficiency). They are making this argument by throwing up a lot of irrelevant (to efficiency) stats about W/L against quads etc. It seemed to me that you were saying a similar thing (the NET is incorrectly or inappropriately measuring team strength).
The complaints about whether using efficiency is actually appropriate for tournament selection I think are well founded. The saving grace though, in my opinion, is that NET actually does not appear to be used this way. It seems to be used much more to gauge the strength of a team's opponents in order to help the selection committee evaluate the (binary) wins and losses on a team's resume. And for this I don't see why efficiency is inappropriate.