AuPM 2.0: The top playoff performers of the databall era

Last year, I tinkered with plus-minus data and the box score to approximate adjusted plus-minus — an augmented plus-minus (AuPM), if you will — which added color to the plus-minus seasons that lack actual play-by-play. In a recent post, I then used that method to approximate playoff only performances, stitching together three-year stretches to create sufficient sample sizes.

But the original AuPM was trained on a relatively small set of seasons and struggled to handle outlying seasons. Top-end campaigns were often shrunk closer to zero in a manner that bothered me, increasing errors at the extremes while nailing most players closer to zero. So I set out to run another regression on a larger data set, and this time traded overall errors for more balance; instead of minimizing the overall error — in which thousands of results fall within two points of neutral — I homogenized the results, keeping error rates more consistent across different strata. The result is a regression that handles high-impact seasons far more accurately. Below is the AuPM 2.0 formula:

AuPM = 0.5543 + Net * 0.1543 + 0.1695 * On  + 0.2632 * BPM – 0.0162 * SumAbove + 0.04242 * (Blocks_per48 + DREB_per48) – 0.0013*(Net * SumAbove)1

As a follow up to Kevin Pelton’s top-50 playoffs by WARP, I thought it would be fun to use AuPM to look at the top postseason players of the databall era (since 1997). This is not a comprehensive analytical ranking like the Backpicks GOAT series, but instead a ranking based on the best players by AuPM only. I tried to balance peak and longevity in listing the most impressive ledgers of the last 22 years. Per game peak AuPM is included in parentheses:

1. LeBron James (+7.9)

The best regular season AuPM using this updated version is Stephen Curry’s legendary 2016 season (+10.7) and, on a per game basis, LeBron James’s iconic 2009 season (+9.9 or +7.8 per game using a basic minutes per game calculation). LeBron’s three-year playoff peak stretched from 2008-10, when he topped out at a nearly identical +7.9 per game. He also holds four of the top-six three-year values on record, and an eye-popping eight of the top-25 marks since 1997. He leads all players with eight entries above +5 and has topped +4 in all 11 three-year stretches of his career.

2. Tim Duncan (+8.1)

This is mostly on the basis of his peak, which was the top value on record (+8.1) from 2001-03. Duncan’s surrounding stretches weren’t far behind — +6.8 from ’02-04 and +6.5 from 1999-02 (he missed the 2000 postseason) — but his best mark after that was “only” +4.1 from 2003-05 before settling into the second phase of his career.

3. Shaquille O’Neal (+6.6)

Shaq’s peak never matched Duncan’s, as he topped out at +6.6 per game during his Finals MVP seasons. He wins on consistency though, as he was in that same ballpark every season from 1998-2004 before dropping off. Given that O’Neal’s regular season AuPM underestimated his actual APM more than Duncan’s, he has a decent argument for the second slot on this list. His seven entries above +4 and five marks above +5 are second only to LeBron.

4. David Robinson (+7.1)

As I mentioned in his GOAT profile, David Robinson’s post-injury career was quietly spectacular. He had mind-boggling impact in the Spurs Twin Towers alongside Duncan, peaking at +7.1 per game in the playoffs from 1998-2000 before adding another good year in 2001.

5. Kevin Garnett (+6.0)

Garnett’s postseason career began as a bumpy affair clouded by low-minute runs, but when he finally logged time he posted the sixth-highest peak on this list at +6.0 per game (with at least 1,000 minutes played). His 2003, 2004 and 2008 playoff stretch clocked in at +5.3, good enough for the seventh-best unique peak with at least 1,500 minutes. He tallied four more stretches above +3 and a massively valuable run for Boston from 2011-13 at +5.2 per game.

6. Michael Jordan (+6.0)

This is more like an honorable mention, as I used Jordan’s ’97 and ’98 season alone to create a 1,697 minute postseason sample that registered an impressive +6.0 AuPM per game.

7. Draymond Green (+5.7)

Green has only logged a few entries in his career, but they’ve all been excellent, owning three of the 33 +5 stretches on record (minimum 1,000 MP). He is a regular season plus-minus darling — his incredible 2016 campaign posted the fourth-highest AuPM per-game value at +7.0 — and his value seems to carry over to the postseason. Only the six players ahead of him on this list and Russell Westbrook have posted better three-year playoff peaks per this metric (with at least 1,500 minutes played).

8. Russell Westbrook (+6.2)

Westbrook owns one of the top playoff peaks in this metric, hitting +6.2 per game from 2015-17 behind his massive box creation numbers. He’s completely outpaced his former MVP teammates Kevin Durant — who has posted a number of playoff stretches just over +3 and peaked this year at +4.2 — and James Harden, who peaked at +4.1 from 2012-14 and regularly posted values close to +4 between 2010 and 2015.

9. Dwyane Wade (+5.2)

Young Wade was spectacular in the playoffs, peaking at +5.2 in 2009 and posting values above +4.5 until 2010. However, his impact dropped alongside LeBron, and his best mark for the rest of his career was a +3.5 from 2010-12.

10. Manu Ginobili (+5.0)

Manu always shines in plus-minus stats, and the playoffs were no exception. He was a major factor in the Spurs third and fourth titles, leading them in on/off and posting AuPM values above +4 from 2003-07, including a peak of +5.0 from 2004-06.

11. Stephen Curry (+5.4)

Curry has played two of the last three postseasons on the heels of a knee injury, and that’s likely knocked his impact down a peg compared to his transcendent regular seasons. Still, his peak from 2013-15 is a solid +5.4, miles ahead of Splash Brother Klay Thompson, whose best playoff mark is +1.6 from 2015-17.

12. Jason Kidd (+4.6)

Like Green, Kidd is a poster child for passing and defense. His regular season AuPM echoes his strong adjusted plus-minus figures — peaking at +5.3 per game — and his playoff stretches aren’t far behind. Kidd’s postseason numbers in Phoenix were subpar in very small samples before posting a slew of high-end seasons in New Jersey and then again in Dallas.

13. Kobe Bryant (+4.4)

Kobe was 13th since 1997 in Pelton’s best playoff rankings by box score, and he falls in the same slot here. Bryant shows great consistency, with four seasons above +4 and six above +3, but never authored a big apex (peaking at +4.4 from 2008-10). Kobe’s teams typically managed without him, and he only led his team in raw on/off three during three playoff stretches of his career.

14. Ray Allen (+6.5)

Allen authored one of 16 three-year stretches over +6, but did so on a hair over 1,000 minutes played. His lack of games played and an ensuing career that peak around +3 make his explosive Milwaukee days look like a flash in the pan, but they are worth noting nonetheless. I split the difference and slotted him among lower peak players.

15. Dirk Nowitzki (+4.7)

In line with his regular season trends, it took Dirk a few years to ramp up his playoff impact. He hit +3 in his first Finals run (2004-06) and peaked at +4.7 in his championship season (2009-11), one of only three stretches in which he led his team in on/off.

16. Paul George (+4.9)

George’s regular season impact metrics have been solid, but he’s been better in the playoffs. (Playoff P!) His on/off numbers were spectacular in Indiana, even if they were likely inflated by lineup rotations that produced huge values for the starting five. Still, George’s values here are worth acknowledgment; he’s one of only eight players to author four stretches with an AuPM of at least +4.

17. Ben Wallace (+4.3)

Big Ben had a brief run, but he was consistently valuable, clocking in at +4.3 over 93 playoff games from 2002-06. This was a slight improvement over his regular season performance in this metric, a testament to his defensive value.

18. Chauncey Billups (+4.4)

The other top Piston during their recent glory years, Billups failed to hit +4 per game in any regular season but peaked slightly higher in the postseason during the second half of the aughts. He’s one of eight players with at least six stretches of +3 or better.

19. Chris Paul (+4.8)

Paul rarely played enough minutes to qualify, but in the stretches that he did, he flirted with high-end impact, peaking at +4.8 from 2014-16 with the Clippers. Another North Carolina-California point guard with small samples could have easily landed here too: Baron Davis. Davis failed to play 1,000 minutes in any three-year window, but posted three consecutive marks right around +5 as a Hornet and Warrior in limited minutes.

20. Pau Gasol (+4.1)

The final spot was a toss up between the Gasol brothers. Pau went over +3 on four separate occasions during his first four Laker seasons, nearly matching Kobe’s value during their reign. His brother, Marc, only authored three qualifying stretches, but posted an impressive +4.9 AuPM from 2011-2013 in nearly 1,400 playoff minutes.

  1. Regression details: Data set was expanded to include Engelmann prior-informed from 2003-2017 and Willard 1998-2000. R-squared was 0.71. Mean Absolute Error (MAE) was 1.11 with a standard deviation of 0.88. MAE of scaled APM values above +5 was 1.56 with a max error of +5.6.

    • For players +4.0 or better, predicted RAPM’s were within 1 of actual RAPM 41 percent of the time, within 2 75 percent of the time and within 3 93 percent of the time.
    • For players between -2.0 and +4.0, predicted RAPM’s were within 1 point 53 percent of the time, within 2 73 percent of the time and within 3 88 percent of the time.

Comments are closed.