RealGM

So I gather this is just a PER that rewards low ast%. Should we just throw PER out and replace it with aper when possible?

Jimmy76 wrote:So I gather this is just a PER that rewards low ast%. Should we just throw PER out and replace it with aper when possible?

I'd like to see some predictive analysis on it, but in terms of theory, it's clearly just an improved version of PER.

Would be very interesting to hear Hollinger's opinion on it.

I'm not sure why we wouldn't go ahead and do it... adjust on actual pace rather than estimating for it. Or am I missing something?

Paydro70 wrote:I'm not sure why we wouldn't go ahead and do it... adjust on actual pace rather than estimating for it. Or am I missing something?

As far as I know, it's adjusting for the actual percentage of his shots that a player is assisted on, rather than going by league average.

kabstah wrote:

Paydro70 wrote:I'm not sure why we wouldn't go ahead and do it... adjust on actual pace rather than estimating for it. Or am I missing something?

As far as I know, it's adjusting for the actual percentage of his shots that a player is assisted on, rather than going by league average.

that and incorporating charges

Paydro70 wrote:I'm not sure why we wouldn't go ahead and do it

I'm guessing by this you mean: Why wouldn't we use the theoretically superior model?

Well, the thing is that with PER, I'd expect Hollinger did some "eyeballing" based on what he felt was intuitively right. Such eyeballing is of course cheating if you're trying to claim a perfect model, but if you're just trying to find something that gives an immediate ballpark, using your intuition is not an unreasonable approach.

Dave Berri, for all his issues, is right when he says that PER will reward some chucking - and I believe APER does that even more in the case of a zero guard. So it's entirely possible that APER would do worse in a test like Levin/Rosenbaum ran on determining the most predictive of the major stats.

However, whenever someone knocks a true point guard for his PER I try to smack them with this. APER is undoubtedly a better measure of an efficient distributor than PER.

Is this what 82games.com uses as well? Or is that a different alteration?

Top 15 by per and aper:
PER/APER
1. Lebron/Lebron
2. Howard/Westbrook
3. Wade/Paul
4. Paul/Wade
5. Love/Nash
6. Durant/Howard
7. Bryant/Bryant
8. Westbrook/Rose
9. Amare/Amare
10. Gasol/Williams
11. Dirk/Love
12. Nash/Durant
13. Rose/Dirk
14. Griffin/Pau
15. Horford/Martin

Basically all the shot creators are getting a boost and big men are taking a hit since they're so often finishers

Doctor MJ wrote:Dave Berri, for all his issues, is right when he says that PER will reward some chucking - and I believe APER does that even more in the case of a zero guard. So it's entirely possible that APER would do worse in a test like Levin/Rosenbaum ran on determining the most predictive of the major stats.

Links?

Jimmy76 wrote:1. Lebron/Lebron
2. Howard/Paul
3. Wade/Westbrook
4. Paul/Paul

Scratch Paul #2. Bump everybody up. Add in Kevin Martin at #15.

Los Soles wrote:

Doctor MJ wrote:Dave Berri, for all his issues, is right when he says that PER will reward some chucking - and I believe APER does that even more in the case of a zero guard. So it's entirely possible that APER would do worse in a test like Levin/Rosenbaum ran on determining the most predictive of the major stats.

Links?

Well here's a link to the Lewin/Rosenbaum paper:

http://www.stumblingonwins.com/LewinRosenbaum2007.pdf

As far as the zero guard statement. The reasoning is that anyone who doesn't get many of his scoring assisted will be rated higher in APER than in PER, and so this will include any primary ball handler. Thus if the such a player shoots inefficiently, him getting more credit for it may make his APER less satisfying to one's sensibilities than his PER.

Would it make sense to increase reward/punishment for inefficiency once you can use actual ast%? Shot creators get their due and you have a harder time chucking yourself to a higher per.

Jimmy76 wrote:Would it make sense to increase reward/punishment for inefficiency once you can use actual ast%? Shot creators get their due and you have a harder time chucking yourself to a higher per.

Depends on what you mean. Such an action I think would produce results that guys like you and me like better, however the PER is derived by actual logic. Any tinkering with logic that Hollinger liked before just to produce better results hurts credibility.

Of course, bad results also hurt your credibility so pick your poison. The reality is that PER is a flawed stat, as is every other stat that relies only on box score numbers (along with with every stat that DOES NOT rely on box score numbers of course, but those stats have a different set of issues).