APBRmetrics

Mountain · Joined: 13 Mar 2007 Posts: 1527

Thanks for the follow-up.

"Define high usage players using a median split (07-08, mp >= 1000). "

19.5 from results at basketball-reference.com for 254 player sample
http://www.basketball-reference.com/fc/tiny.cgi?id=L1457

"Within that group define high and low ORTG players with another median split."

Per b-r 108.8.

"The correlation b/t offensive apm and usage is the same across efficiency levels (r = .43)."

Not sure what you mean by this, especially the "b/t". Using your choices (which would be?) you already this?

Brian M · Joined: 25 Nov 2006 Posts: 40

b/t = between, sorry.

I mean that within the high usage, low efficiency group, the correlation between offensive apm and usage is .43. In the high usage, high efficiency group, the correlation between offensive apm and usage is also .43. So the relationship between usage and offensive apm is not modulated by ORTG, as also suggested by the weak interaction effect in the regression analysis.

Mountain · Joined: 13 Mar 2007 Posts: 1527

Ok on the b/t.

So you defined high/low usage and efficiency previously to answer that issue? How did your values compare to these medians I looked up? The same?

**Ilardi** · Joined: 15 May 2008 Posts: 263 Location: Lawrence, KS

Mike G · Posted: Wed Nov 12, 2008 6:21 pm Post subject:

**Ilardi** · Joined: 15 May 2008 Posts: 263 Location: Lawrence, KS

Neil Paine · Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA

So Steve, are you saying that a team's efficiency differential is not equal to the minute-weighted average of its players' adjusted +/- scores, but rather the minute-weighted sum? This is something I've always wondered and I've seen comments that suggest both, so I'm glad someone is willing to set the record straight.

**Ilardi** · Joined: 15 May 2008 Posts: 263 Location: Lawrence, KS

Mike G · Posted: Wed Nov 12, 2008 9:09 pm Post subject:

Brian M · Joined: 25 Nov 2006 Posts: 40

deepak · Joined: 26 Apr 2006 Posts: 664

Neil Paine · Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA

Mountain · Joined: 13 Mar 2007 Posts: 1527

Adjusted +/- is done in the context of teams with roles to some degree.

Using Brian's useful file what can we note about the usage part of context?

With a few exceptions guys who achieved a +5 or better on offensive +/- had a 20+% usage rate. The exceptions were Jason Kidd, Chris Quinn and Peja barely while the other 11 had it and average was a 24.3 usage rate.

51 players over 1000 minutes had a 24.3+ usage rate so under 22% achieved this +5 adjusted +/- goal. The average is about +2.5 offensive adjusted with a 26.7 usage rate. If you are yielding less on adjusted +/- with this level of usage you should be evaluated very carefully.

You could look at individual adjusted/usage ratios. The average for this group was about offensive adjusted /usage= 9%. The top guys were Paul (at 36%), Bryant (28.5%), James (27.5%) then Garnett (26%).

Iverson is at a fine 19%. Yao was a zero.

The lowest in this group were Nick Young (-17%), Jermaine O'Neal, David West, Charlie V. and Durant (-5%).

The average defensive adjusted is a -0.2 so proibably half or more of the guys are offensively biased. 2 way players are the way to go as much as possible.

5 teams had 3 guys in this group, 9 more had 2. So remaining teams had about 1 each. With one exception all the guys who were negative were on teams with multiples such players and maybe that is a factor- team inability to optimize offensive adjusted for those 2. But others manage to do it ok.

The next 60 guys on usage return almost no offensive adjusted / usage ratio, at just a tiny fraction above 0. Should secondary scorers by this criteria yielding nothing on offensive adjusted on average should make offense lower priority or a more frantic hunt?

The best in the group are Devin Haris, Jason Terry, Billups and Raashard Lewis and they are the few standouts. The very worst are Curry, Kaman, Warrick, Nachbar and surprising to me Farmar. Stuckey at -10%.

The average defensive adjusted is -0.77. Seems to me you'd want to get better than group average on defense wherever possible even if they weren't as prominent a choice from perception of their offensive ability. Again 2 way players have 2 chances to add up to a good contribution worthy of a moderate or bigger salary.

Of the 108 between 15-20% usage the average adjusted offensive return ratio is -5%. If you can find a guy who is neutral on offensive adjusted to be your third guy or better you've done well. Jason Kidd, Sasha Vujacic, Chris Quinn and Vlade Radmanovic are the runaway kings of the third tier.

Among the every worst are Brandon Bass,, Jacque Vaughn Odom, Yi, Pavolovic Haslem, Rondo.Blatche, Jack, Blount, Cook and Mohammed. The average defensive adjusted -0.15. So on average they aren't hurting much on D.

The best combo of offensive and defensive adjusted up to zero or occasionally better is the goal. Some of the best overall are Mike Miller, Peja, Tony Allen, Kidd, Brewer, B Miller, T Young, KMart, Songaila, Dickau, Kirilenko, Maxiell (Utah knows this tier) Radmanovic, Chandler and Moon.

Mountain · Joined: 13 Mar 2007 Posts: 1527

The APM values are per 100 possessions at basketballvalue and the most recent multi-year study. Steve's 06-07 dataset was per 40 minutes. Dan Rosenbaum and Dave Lewin's prior work was per 100 possessions.

I did the math on last season's Sonics and it works as a sum.
Durant cost the team -500 some points during his season's actual possessions.

Mike G · Posted: Thu Nov 13, 2008 6:40 am Post subject: