APBRmetrics

I started with the top 100 on BPM for last season who played 25 plus minutes per game to get a manageable initial data set.

I then compiled PER, ORtg, DRtg, OBPM, DBPM, DRE, ORPM and DRPM and standardized them by st. dev. of the group around its mean.

Then, as an initial blend I found a Blended metric value as following: .1 PER + .05 ORtg +.05 DRtg +.1 OBPM +.1 DBPM +.2 DRE +.15 ORPM +.25 DRPM= Blended metric value. I may tweak the formula later. DRPM got more value to make a crude attempt to make up for the lack of shot defense or lack of individualized shot defense in other metrics. A blend of 60% boxscore based metrics and 40% adjusted plus minus seemed like a decent balance to start with.

But, fwiw, here is my Blended metric 1.0 (pseudo) top 100:

Player Blend
Anthony Davis 2.86
Stephen Curry 2.74
Kawhi Leonard 2.27
LeBron James 2.22
James Harden 2.17
Chris Paul 2.16
Russell Westbrook 1.89
Draymond Green 1.74
Tim Duncan 1.71
Kevin Durant 1.49
DeMarcus Cousins 1.43
Tyson Chandler 1.41
DeAndre Jordan 1.38
Paul Millsap 1.28
Danny Green 1.26
Khris Middleton 1.23
Tony Allen 1.20
Rudy Gobert 1.20
Jimmy Butler 1.18
Marc Gasol 1.04
Kyle Korver 1.01
George Hill 0.98
Gordon Hayward 0.92
John Wall 0.90
Blake Griffin 0.87
Al Horford 0.85
Wesley Matthews 0.83
Kyle Lowry 0.82
Zach Randolph 0.81
Damian Lillard 0.76
Klay Thompson 0.73
Greg Monroe 0.68
Eric Bledsoe 0.67
Derrick Favors 0.65
Kyrie Irving 0.64
Nikola Vucevic 0.64
Marcin Gortat 0.63
Pau Gasol 0.62
Jrue Holiday 0.61
Serge Ibaka 0.60
LaMarcus Aldridge 0.58
Jeff Teague 0.57
Kevin Love 0.52
Mike Conley 0.33
Chandler Parsons 0.31
Matt Barnes 0.28
Mike Dunleavy 0.28
Darren Collison 0.26
Lou Williams 0.23
Amir Johnson 0.21
Joakim Noah 0.21
Kenneth Faried 0.21
Trevor Ariza 0.20
Luol Deng 0.17
Tyreke Evans 0.16
Deron Williams 0.15
Carmelo Anthony 0.14
Jared Sullinger 0.09
J.J. Redick 0.09
DeMarre Carroll 0.06
Marcus Smart 0.02
Robin Lopez 0.02
Isaiah Thomas 0.02
Brandon Jennings 0.01
Robert Covington 0.00
Terrence Jones -0.02
Gorgui Dieng -0.03
Brook Lopez -0.03
Ty Lawson -0.04
Bradley Beal -0.05
Monta Ellis -0.06
Andre Iguodala -0.06
Giannis Antetokounmpo -0.07
Chris Bosh -0.08
Harrison Barnes -0.10
Dirk Nowitzki -0.13
Paul Pierce -0.14
Nerlens Noel -0.15
Patrick Patterson -0.16
Nicolas Batum -0.20
Marvin Williams -0.21
David West -0.23
J.R. Smith -0.23
Rudy Gay -0.25
Al Jefferson -0.25
Reggie Jackson -0.27
Goran Dragic -0.31
P.J. Tucker -0.35
Taj Gibson -0.44
Steven Adams -0.45
Dwyane Wade -0.45
Joe Johnson -0.47
Kemba Walker -0.49
Josh Smith -0.50
Victor Oladipo -0.59
Brandon Knight -0.70
Courtney Lee -0.70
Patrick Beverley -0.76
Wilson Chandler -0.96
Kobe Bryant -1.22


Note: If I looked at everyone, instead of starting with BPM top 100, some might drop out and some come in. That is also for a later time, perhaps.

Ive thought about doing something similar to gather an all-time list of players, seasons, etc. Especially a blended metric for Guards, who dont always get their just due in some metrics, and the value of them differs by a wide margin in some metrics..

I didn't do it this way for this post, but for a more serious comparison effort I would indeed compare PGs to PGs only, wings to wings, bigs to bigs and perhaps customize the metric to the position's core responsibilities to some degree.

For instance, AST%, TOV%, PPR for PG's.
TS%, 3PAr, FTr/FGA for Wings, and BLK%, REB%, (ORB%, DRB%, as well), perhaps Charges drawn for Bigs.

This made me wonder how well each of Crow's inputs correlate with the final blended metric.
But he uses RPM, and I don't know an easy way to grab those. So I made my own version with stats at b-r.com.
The blended metric is called xyz (as usual) for lack of a better name. It actually maximizes the total (sum of) correlations between xyz and each component of the weighted sum.

The weight (coefficient) given each number, and the correlation with xyz:Sta% = GS/G, max = 1.00
Other rates are as listed at b-r.com, i.e., per 100 mostly.
FTr, 3PAr, and WS/48 are in .000 format, thus their larger coefficients.
Correlations are for 275 players with >1000 minutes this season.

xyz top 40 for 2014-15:

Mind if I borrow this blended metric for some comparisons/analysis?
Perhaps the 1,647,428th comparison of MJ & LBJ ever , or Magic & Bird, or Magic & LBJ.
I'd love to see a regression like this, using AST% and TOV% also...

Go ahead.



For those who think the DRPM weight is too high, note that PER and DRE have no shot defense at all and are 30% of the blend. The extra 10% of DRPM might be a little low to fully backfill for that but that's where I start it, knowing anything more would turn-off some possible users even more.