APBRmetrics

The only trend I see so far in the under and overrated compared to RAPM data so far, it that the over and under rating of these cases is consistent on both sides of court except for underrated centers. Seeing individual stats might help but good chance it doesn't. Maybe seeing overall performance and discrete stats as player / team ratios might help and avg career sub stats as ratios too might help. Seems more likely to me that these cases are a result of context difference than individual performance difference but hard to guess. Perhaps rapm error is the main difference. Perhaps the BPM adjustment? Is there a pattern to how that influences these player ratings? Is there a pattern to how these players actually did when way ahead / behind vs the assumptions? Is there a notable difference to how often they were one or the other?

I dunno if it means something or not but 70-75% of the time the O & D numbers are further from 0 on RAPM than BPM.

This season's BPM columns are blank at b-r.com

Mike G wrote:This season's BPM columns are blank at b-r.com

?

They show up for me.

http://www.basketball-reference.com/lea ... anced.html

Of course BPM will have some holes if it's only using box score stats. I mean, think about rebounding and how misleading those numbers can be, and there's nothing in the box score to help with that. You'd have to go deep into player tracking stats like boxouts and contested boards to figure that out. It's also going to underrate good fundamental defenders who don't pick up a lot of blocks and steals, from Battier to Robin Lopez.

Also, look at these underrated PFs:
LaMarcus Aldridge ║ 4 ║ 20460 ║ 20.0 ║ 0.143 ║ 5.8 ║ 2.3 ║ 3.4 ║ 1.2 ║ 0.9 ║ 0.3 ║ -4.6 ║ -1.4 ║ -3.1 ║
║ Kevin Garnett ║ 4 ║ 35362 ║ 24.0 ║ 0.204 ║ 9.7 ║ 2.4 ║ 7.3 ║ 6.0 ║ 2.4 ║ 3.6 ║ -3.7 ║ 0.0 ║ -3.7 ║
║ Dirk Nowitzki ║ 4 ║ 38661 ║ 24.2 ║ 0.218 ║ 7.4 ║ 5.2 ║ 2.1 ║ 4.0 ║ 3.5 ║ 0.5 ║ -3.4 ║ -1.7 ║ -1.6 ║
║ Amir Johnson ║ 4 ║ 11193 ║ 16.3 ║ 0.146 ║ 5.8 ║ 1.7 ║ 4.1 ║ 2.6 ║ 0.3 ║ 2.3 ║ -3.2 ║ -1.4 ║ -1.8 ║
║ Rasheed Wallace ║ 4 ║ 25408 ║ 17.4 ║ 0.141 ║ 5.3 ║ 1.0 ║ 4.3 ║ 2.4 ║ 0.7 ║ 1.6 ║ -2.9 ║ -0.3 ║ -2.7 ║
║ Jason Thompson ║ 4 ║ 12336 ║ 14.0 ║ 0.077 ║ 0.9 ║ 0.7 ║ 0.2 ║ -1.7 ║ -1.4 ║ -0.3 ║ -2.6 ║ -2.1 ║ -0.5 ║

That pattern to me is obvious in at least two factors, with the possible exception of Jason Thompson (although he sorta counts for one of the factors.)

AcrossTheCourt wrote:Of course BPM will have some holes if it's only using box score stats. I mean, think about rebounding and how misleading those numbers can be, and there's nothing in the box score to help with that. You'd have to go deep into player tracking stats like boxouts and contested boards to figure that out. It's also going to underrate good fundamental defenders who don't pick up a lot of blocks and steals, from Battier to Robin Lopez.

Also, look at these underrated PFs:
LaMarcus Aldridge ║ 4 ║ 20460 ║ 20.0 ║ 0.143 ║ 5.8 ║ 2.3 ║ 3.4 ║ 1.2 ║ 0.9 ║ 0.3 ║ -4.6 ║ -1.4 ║ -3.1 ║
║ Kevin Garnett ║ 4 ║ 35362 ║ 24.0 ║ 0.204 ║ 9.7 ║ 2.4 ║ 7.3 ║ 6.0 ║ 2.4 ║ 3.6 ║ -3.7 ║ 0.0 ║ -3.7 ║
║ Dirk Nowitzki ║ 4 ║ 38661 ║ 24.2 ║ 0.218 ║ 7.4 ║ 5.2 ║ 2.1 ║ 4.0 ║ 3.5 ║ 0.5 ║ -3.4 ║ -1.7 ║ -1.6 ║
║ Amir Johnson ║ 4 ║ 11193 ║ 16.3 ║ 0.146 ║ 5.8 ║ 1.7 ║ 4.1 ║ 2.6 ║ 0.3 ║ 2.3 ║ -3.2 ║ -1.4 ║ -1.8 ║
║ Rasheed Wallace ║ 4 ║ 25408 ║ 17.4 ║ 0.141 ║ 5.3 ║ 1.0 ║ 4.3 ║ 2.4 ║ 0.7 ║ 1.6 ║ -2.9 ║ -0.3 ║ -2.7 ║
║ Jason Thompson ║ 4 ║ 12336 ║ 14.0 ║ 0.077 ║ 0.9 ║ 0.7 ║ 0.2 ║ -1.7 ║ -1.4 ║ -0.3 ║ -2.6 ║ -2.1 ║ -0.5 ║

That pattern to me is obvious in at least two factors, with the possible exception of Jason Thompson (although he sorta counts for one of the factors.)

Decided to take a quick look at the top of BPM career results (regular season '79-80 to '13-14) compared to my WAR/48 results from the time.

BPM rank: http://www.basketball-reference.com/pla ... der_by=bpm

My WAR rank (WAR/48 & rank in purple after TS%) http://hoopsnerd.com/wp-content/uploads ... nkings.pdf

Top 25 BPM:

#1 & #2 for both, LeBron then MJ.
#3 for BPM is Charles Barkley (he is 10th WAR/48)
#4 for BPM, Larry Bird (7th WAR/48)
#5 for BPM, Magic (6th in WAR/48)
#6 BPM, CP3 (5th in WAR/48)
#7 BPM, David Robinson (#3 WAR/48)
#8 BPM, Drexler (26th WAR/48)
#9 BPM, Wade (11th WAR/48)
#10 BPM, Ginobli (27th WAR/48)
#11 BPM, Duncan (8th WAR/48)
#12 BPM, Dr. J (16th WAR/48)
#13 BPM, Garnett (14th WAR/48)
#14 BPM, Karl Malone (12th WAR/48)
#15 BPM, Kirilenko (96th WAR/48)
#16 BPM, Pippen (64th WAR/48)
#17 BPM, Shaq (#4 WAR/48)
#18 BPM, Kawhi Leonard (#92 WAR/48)
#19 BPM, Hakeem (#9 WAR/48)
#20 BPM, McGrady (#28 WAR/48)
#21 BPM, Noah (#53 WAR/48)
#22 BPM, Kobe (#17 WAR/48)
#23 BPM, Kidd (#51 WAR/48)
#24 BPM, Durant (#13 WAR/48)
#25 BPM, Harden (#43 WAR/48)

Top 25 WAR/48 not listed above:
#38 BPM, Dirk (#15 WAR/48)
#78 BPM, Dwight Howard (#18 WAR/48)
#147 BPM, Patrick Ewing (#19 WAR/48)
#141 BPM, Yao Ming (#20 WAR/48) (1.8)
#55 BPM, John Stockton (#21 WAR/48)
#153 BPM, Moses Malone (#22 WAR/48)
#51 BPM, Kareem (#23 WAR/48)
#28 BPM, Blake Griffin (#24 WAR/48)
#207 BPM, Alonzo Mourning (#25 WAR/48)

Just want to point out - according to BPM - Kirilenko, when he played, appears to have about 2.5 times the value of Ewing, Yao Ming, or Moses Malone. Almost 4 times the Value of Alonzo Mourning. Kirilenko's career average BPM is better than ANY single year Alonzo Mourning, Moses Malone, Yao Ming, Dikembe Mutombo, or even Robert Parish logged. Kirilenko has NINE seasons of BPM better than the very best Yao Ming season. If only those other big guys got got just a little more assists per game tied to their great rebounding, they could have been decent.

If memory serves, Winston-Sagarin many years ago were much maligned because they ended up with results on a season that had Kirilenko ranked #1 over Shaq. "Laugh Test" became a term. Now, we create a box score metric that best mimics Winston-Sagarin, has Kirilenko as better than Shaq, & it's all cool. This confuses me. Dave Berri is a running joke around here because of his weird results - I know, his correlations suck. But, shouldn't we want both - a metric that best fits the eye test of many die hard fans & correlates well also, even if it correlates slightly worse than BPM?

Don't get me going on Bo Outlaw being MUCH better than many Hall of Famers according to BPM.

There's WAY too much "noise" to me with BPM - I think particularly around tying assists & rebounding together. Two statistically identical players, except one gets 10 rpg & 1 apg, the other gets 7 RPG & 3 apg - I'm guessing the first by BPM could be slightly above average while the 2nd a HoF'er. That's WAY to much a shift for me tied to a couple stats - even if in an overall sense they seem to correlate a little better year to year. Tying stats together when doing something like this seems dangerous anyway - asking for weird results.

For fun, grew up a Pistons fan, 3 ex Bad Boys who always seem "tricky" to evaluate by any metric:
#112 BPM, Isiah Thomas (#79 WAR/48)
#50 BPM, (#154 WAR/48)
#387 BPM, Joe Dumars (#253 WAR/48)

Eye test pattern identification has its perils but in the bpm / war comparison aren't most of the big differences wings who pass very well (and bpm gives extra credit), bigs who don't pass that much relative to other positions (bpm overall rank suffers?) and pgs who reb. above or below avg. (affecting bpm compared to war)?

Well for what it's worth, Bo Outlaw did really well in the late 90's/early 00's "pure" RAPM. He does appear to be an underrated player via his defense.

Kirilenko has some historically crazy block/steal numbers in his prime and a pretty good amount of assists too, so unless you have better defensive stats it's hard to get around that issue. Shaq, however, gets most of his value on offense, but it shouldn't be too difficult to measure that.

By the way, we shouldn't dismiss results just because they don't fit our preconceived notions. (I've always felt Moses was overrated even before these numbers, even if others saw him as a true MVP-level guy.)

edit: I just thought of the problem with Kirilenko's rating. Blocks are a pretty mediocre measure of defense. And some blocks are more valuable than others. I think Kirilenko got more of his block shots on the perimeter, which they are generally not worth as much as near the rim (also because I believe those blocks are picked up by the defense less often.) From his blocked shots, he looks like a great rim protector like Mutombo or Mourning were -- but perhaps he's not, more of a weakside/perimeter shot blocker.

Best seasons by VORP by players who were not selected to All-Star team (since 1974)That first column, xy1 = sqrt(WS*VORP)
Sorting by that, we get about 5 new player-seasons in the top 20.From top 100 seasons (for non allstars) http://bkref.com/tiny/9KIk4
... we still have 84 after removing those from 1999 (no allstar selections).
Appearing 3 times in the list: Cedric Maxwell, Drexler, Brand, Kidd, Nance, Ginobili, Blaylock, Unseld.
With 2 appearances: Divac, Pressey, Lever, Iguodala, Kirilenko, Anthony Mason, Ben Wallace, Bo Outlaw

1997 had 6 high-VORP players who missed allstar selection.

AcrossTheCourt wrote:Well for what it's worth, Bo Outlaw did really well in the late 90's/early 00's "pure" RAPM. He does appear to be an underrated player via his defense.

Kirilenko has some historically crazy block/steal numbers in his prime and a pretty good amount of assists too, so unless you have better defensive stats it's hard to get around that issue. Shaq, however, gets most of his value on offense, but it shouldn't be too difficult to measure that.

By the way, we shouldn't dismiss results just because they don't fit our preconceived notions. (I've always felt Moses was overrated even before these numbers, even if others saw him as a true MVP-level guy.)

edit: I just thought of the problem with Kirilenko's rating. Blocks are a pretty mediocre measure of defense. And some blocks are more valuable than others. I think Kirilenko got more of his block shots on the perimeter, which they are generally not worth as much as near the rim (also because I believe those blocks are picked up by the defense less often.) From his blocked shots, he looks like a great rim protector like Mutombo or Mourning were -- but perhaps he's not, more of a weakside/perimeter shot blocker.

For what it's worth, BPM actually weights blocks less than most other stats. It values steals very highly, and also counts ast*rebounds as a positive on defense (apparently such an interaction correlates well with players that are good on defense--aware and athletic).

Blocks are a pretty mediocre measure of defense.

on the contrary for a single statistic measure blocks are a very good indicator of team defense:

how about looking at all team final regular season annual stats since 1977-78 (38 years of regular season team data, 1028 team seasons), and comparing team bs/100poss to team eFG% allowed and team defensive efficiency (pts/poss) allowed:

----bs-------allowed----allowed
/100poss---avg-eFG%-avg-pts/poss
8.0+----------.474-------1.036
7.5-8.0------.469-------1.024
7.0-7.5------.471-------1.029
6.5-7.0------.479-------1.036
6.0-6.5------.481-------1.042
5.5-6.0------.483-------1.042
5.0-5.5------.488-------1.048
4.5-5.0------.490-------1.049
4.0-4.5------.499-------1.064
3.5-4.0------.501-------1.063
3.0-3.5------.504-------1.077
<3.0----------.519-------1.101

there are exceptions to every correlation on a team-by-team single season basis, but the fact is that overall higher blocks do track lower eFG%s allowed and lower pts/poss allowed...

DSMok1 wrote:For what it's worth, BPM actually weights blocks less than most other stats. It values steals very highly, and also counts ast*rebounds as a positive on defense (apparently such an interaction correlates well with players that are good on defense--aware and athletic).

In my tests, the regression screamed at me to include BLK% and AST% with BPM. However I couldn't overcome overfitting. I definetely think BPM undervalues blocks and assists(ast*reb is the problem imo).

DSMok1 wrote:and also counts ast*rebounds as a positive on defense (apparently such an interaction correlates well with players that are good on defense--aware and athletic).

Yeah, Dikembe Mutombo & Alonzo Mourning couldn't hold a candle defensively to Sir Charles & Larry Bird. It's because they don't pass.

bchaikin wrote:Blocks are a pretty mediocre measure of defense.

on the contrary for a single statistic measure blocks are a very good indicator of team defense:

how about looking at all team final regular season annual stats since 1977-78 (38 years of regular season team data, 1028 team seasons), and comparing team bs/100poss to team eFG% allowed and team defensive efficiency (pts/poss) allowed:

----bs-------allowed----allowed
/100poss---avg-eFG%-avg-pts/poss
8.0+----------.474-------1.036
7.5-8.0------.469-------1.024
7.0-7.5------.471-------1.029
6.5-7.0------.479-------1.036
6.0-6.5------.481-------1.042
5.5-6.0------.483-------1.042
5.0-5.5------.488-------1.048
4.5-5.0------.490-------1.049
4.0-4.5------.499-------1.064
3.5-4.0------.501-------1.063
3.0-3.5------.504-------1.077
<3.0----------.519-------1.101

there are exceptions to every correlation on a team-by-team single season basis, but the fact is that overall higher blocks do track lower eFG%s allowed and lower pts/poss allowed...

No.

Team stats!=individual stats, or else we'd ignore offensive rebounding for players.

If you include SportVU stats like the rim protection ones, distance while defending, etc. blocks are a lot less valuable, drop out completely, or actually end up with a negative correlation. Blocks are not a good measure of defense, especially now with all the information we have.

Statman wrote:

DSMok1 wrote:and also counts ast*rebounds as a positive on defense (apparently such an interaction correlates well with players that are good on defense--aware and athletic).

Yeah, Dikembe Mutombo & Alonzo Mourning couldn't hold a candle defensively to Sir Charles & Larry Bird. It's because they don't pass.

I agree that the results appear somewhat questionable in a few spots, but I trust the process and I'm not going to manipulate the results just because I don't like them.

BLK% alone has an R^2 of 0.25 onto Defensive RAPM, which is the best individual stat's correlation, but once other stats are added, its weight actually becomes rather small.

Kevin Garnett had easily the best defensive RAPM over the 14 year dataset, and his block rate wasn't particularly high. Tim Duncan was #2, and his block rate was higher, but still not elite. I'm sure that has some influence on the regression. Javale McGee and Serge Ibaka had extremely high block rates, and don't rate that well in DRAPM.

Interestingly, Sqrt(BLK%*AST%) has the same R^2 on Defensive RAPM as pure BLK% does.