Predicting NBA Playoffs using RAPM (updated with Finals)

[colts18]

I updated the calculations using regular season team numbers to regress the projections. Regular season adjusted net efficiency is 1/3 (postseason 2/3) of the projections.

Projected team SRS (not per 100):
Spurs 12.2
Blazers 5.3

Thunder 8.5
Clippers 10.0

Pacers 5.3
Wizards 2.6

Heat 7.6
Nets 2.0

Based on that, here are the series win probability:
Spurs 92.0
Blazers

Thunder 44.9
Clippers

Pacers 73.7
Wizards

Heat 86.9
Nets

Not too much of a difference though the Spurs rating is the one that takes the biggest hit when including regular season numbers despite the fact that they had an 8 SRS this season. Thunder get a slight boost even though the Clippers had a better regular season SRS than them.

[colts18]

The loss of Ibaka is huge for the Thunder. If I split his minutes evenly between Perkins, Collison, and Adams, the Thunder go from a +9.9 per 100 team to a +6.4 per 100 team. xRAPM loves Ibaka, but it hates Steven Adams and Perkins just as much. Both of them are -4 players in that rating while Collison is a +6.


Team SRS ratings:
Spurs 13.8
Thunder 6.0 (with Ibaka getting 0 minutes)

Pacers 6.3
Heat 10.5


Series odds:
Spurs 93.9%
Thunder

Pacers 41.0% (this includes game 1)
Heat

The Pacers were at 27.6% chance of winning the series before game 1. xRAPM still loves the Spurs a lot. The Spurs top 7 minute getters are all +1.5 or better in xRAPM. The Spurs have 3 players who are negative xRAPM players. 89% of their playoff minutes have come from positive players. Both the Heat and Thunder have 68.9% of their minutes coming from positive players. The Pacers have 73.5% of their minutes coming from positive players. The pacers are an interesting case because all 5 of their starters are positive but every member of their bench is negative.

[J.E.]

The pacers are an interesting case because all 5 of their starters are positive but every member of their bench is negative.

Yeah, the Granger for Turner deal isn't looking too hot right now, and neither does Scola+pick for Green and Plumlee. Scola just looks awful whenever I see him play

The Spurs are currently favored by 6 at home for game 1, hinting at a 6-3.5 = 2.5 difference in strength

[colts18]

Yeah, the Granger for Turner deal isn't looking too hot right now, and neither does Scola+pick for Green and Plumlee. Scola just looks awful whenever I see him play

The Spurs are currently favored by 6 at home for game 1, hinting at a 6-3.5 = 2.5 difference in strength

I would be interested to see if the Spurs would be considered favorites against the Heat. Based on xRAPM and regular season RAPM, the Spurs are the clear favorites. It could be a good opportunity for an analytical guy to make money by betting against the public favorite.

[AcrossTheCourt]

What value are you using for HCA? And where did you get that value?

[colts18]

What value are you using for HCA? And where did you get that value?

3.5 points. HCA could be even worth 4+ points

viewtopic.php?f=2&t=8021&p=13098

[Mike G]

The loss of Ibaka is huge for the Thunder. If I split his minutes evenly between Perkins, Collison, and Adams, the Thunder go from a +9.9 per 100 team to a +6.4 per 100 team. xRAPM loves Ibaka, but it hates Steven Adams and Perkins just as much. Both of them are -4 players in that rating while Collison is a +6.
...

So who has gotten Ibaka's minutes? Two games here is not a representative sample, with so many garbage minutes, but:

Code: Select all

.minutes / G      prev   WCF    diff
Perry Jones        1.2   13.0   11.8
Derek Fisher      13.6   24.0   10.4
Steven Adams      15.1   25.0    9.9
Jeremy Lamb        1.5   10.5    9.0
Andre Roberson      .0    4.5    4.5
Nick Collison     10.3   14.5    4.2
Reggie Jackson    26.7   27.5     .8
Hasheem Thabeet     .3     .0    -.3
Caron Butler      24.5   22.5   -2.0
Kendrick Perkins  21.0   19.0   -2.0
Thabo Sefolosha   15.5   13.0   -2.5
Russell Westbrook 39.5   32.5   -7.0
Kevin Durant      44.5   35.0   -9.5
Serge Ibaka       33.9     .0  -33.9

Brooks has declined to use Thabeet. Collison, Butler, and Perkins are a net zero in added minutes.
Guard minutes, led by Fisher and Lamb, are up 11 mpg. That's a lot of 3-guard-lineup time.

[colts18]

The last 2 games seem to be vindication for xRAPM. The Spurs have been +28 in this series and +10 for the playoffs. The Spurs have basically the same O rating as the Heat for the playoffs (114 O Rating), but their defense has been about -4 points better. I would put them as the favorites over the Heat at this point.

[xkonk]

The last 2 games seem to be vindication for xRAPM.

In what sense? I don't think anyone had the Spurs winning by 17 and 35. Does xRAPM predict a bigger Spurs margin than other systems? Had they won by 1 and 3, or managed to lose one of the games, would these two games be condemning evidence against xRAPM?

[Mike G]

In the 2005 Finals, the Spurs won games 1 & 2 at home, by 15 and 21.
The Pistons won the next 2, by 17 and 31.
In the end, Det outscored SA by 1.9 PPG.

[colts18]

The last 2 games seem to be vindication for xRAPM.

In what sense? I don't think anyone had the Spurs winning by 17 and 35. Does xRAPM predict a bigger Spurs margin than other systems? Had they won by 1 and 3, or managed to lose one of the games, would these two games be condemning evidence against xRAPM?

Good point. I should have said these playoffs vindicated xRAPM for giving the Spurs a huge rating. the Spurs seemed like the outlier in xRAPM. I imagine their xRAPM rating would have been similar to all-time teams like the 96 Bulls or 87 Lakers. But we should probably wait until the end of the playoffs. If they continue this vs the Thunder then have a good series vs the Heat, then maybe we should have a discussion on the 2014 Spurs being an all-time team.

[colts18]

xRAPM for the finals:

SRS values; Off, def, total

using playoff minutes from all rounds
Spurs 5.8 7.6 13.3
Heat 6.8 2.8 9.6

using only minutes from Conference finals
Spurs 5.6 6.7 12.3
Heat 6.0 2.6 8.6

Win/Loss% on a neutral court:
Spurs 0.637
Heat

Spurs 0.633
Heat

Series win/loss%:
Spurs 0.803
Heat

Spurs 0.798
Heat

I've seen Vegas have the Spurs as -130 favorites and another book at even odds. B-R has the Heat at 55% chance of winning. But xRAPM still loves the Spurs a lot. It has the Spurs as an 80% favorite. 9 out of the 10 Spurs players who have played 100+ minutes in the postseason are a positive xRAPM. Only 7 out of 11 Heat with 100+ MP have a positive xRAPM.

[colts18]

Using minutes from the conference finals.

Spurs in 4- 14.8%
Heat in 4- 1.6%

Spurs in 5- 28.1%
Heat in 5- 3.6%

Spurs in 6- 17.8%
Heat in 6- 6.5%

Spurs in 7- 19.5%
Heat in 7- 6.5%

Spurs 79.8% series win%
Heat 20.2% series win%

Series on average will last 5.6 games. 48.1% of the time it will go under 6 games, 51.9% over 6 games. 25.6% chance this series goes 7 like it did last year.

[J.E.]

colts18, 'SRS' is a team rating system by itself ('Simple Rating System') which doesn't use any player specific information (see here)

It seems the lines have moved from Spurs -3.5 for the first game and being a very very slight favorite for the series to -4.5 and being a 60% favorite for the series. I'm guessing a lot of it will depend on Tony Parker's ability to play (well)

colts18, you're using RPM from ESPN, right? I haven't updated xRAPM on my site for a while

[colts18]

colts18, 'SRS' is a team rating system by itself ('Simple Rating System') which doesn't use any player specific information (see here)

It seems the lines have moved from Spurs -3.5 for the first game and being a very very slight favorite for the series to -4.5 and being a 60% favorite for the series. I'm guessing a lot of it will depend on Tony Parker's ability to play (well)

colts18, you're using RPM from ESPN, right? I haven't updated xRAPM on my site for a while

I'm using RPM from the ESPN site because it updates in the playoffs. By SRS I mean RPM values multiplied by 0.939 (# of possessions in average game) to get a per game SRS value.