web analytics

NBA Final Regular Season Bayesian Ratings

April 15, 2011
By

There is a new page on this site!  I’m putting together an NBA Playoffs preview; this is the first of the data for that:

Here is a table of playoff bound teams and their season-long adjusted efficiency ratings. Ratings are adjusted for location, pace (obviously, since these are per 100 possessions), opponent (recursively) and rest day situation. Previous is the rating of the team at the time of the trade deadline.

Rank Chg Team Off Eff Def Eff Eff. Margin Change Prev SoS Pace
1 MIA 4.35 -3.06 7.41 -0.63 8.04 -0.79 90.3
2 3 CHI 0.19 -7.08 7.27 1.57 5.71 -0.73 89.7
3 1 LAL 4.01 -2.64 6.65 0.12 6.52 -0.05 90.0
4 -2 SAS 4.13 -1.84 5.98 -1.05 7.03 -0.22 91.6
5 -2 BOS -0.96 -6.41 5.44 -1.53 6.97 -0.46 89.4
6 ORL 0.87 -4.49 5.36 -0.17 5.54 -0.54 90.7
7 1 DEN 4.96 0.00 4.96 1.94 3.02 -0.04 95.1
8 -1 DAL 1.50 -3.12 4.63 1.04 3.58 -0.07 90.6
9 2 OKC 3.95 0.28 3.67 1.58 2.09 -0.33 91.9
10 2 MEM 0.23 -2.47 2.71 0.86 1.85 0.21 91.0
11 4 HOU 3.52 1.18 2.33 1.68 0.66 0.03 93.4
12 1 POR 1.83 -0.49 2.32 1.09 1.23 0.62 86.9
13 -3 PHI -0.60 -1.93 1.34 -0.87 2.20 -0.26 90.9
14 -5 NOH -0.99 -2.10 1.11 -1.28 2.40 0.11 88.7
15 -1 NYK 3.92 2.92 1.00 0.21 0.79 0.20 95.3
16 2 PHO 2.65 3.22 -0.57 -0.09 -0.48 0.33 93.7
17 -1 ATL -0.67 0.33 -1.01 -1.45 0.44 -0.11 88.5
18 2 MIL -6.60 -5.22 -1.39 0.05 -1.44 -0.49 89.0
19 -2 IND -2.12 -0.72 -1.40 -1.10 -0.30 -0.20 94.5
20 -1 UTA 1.09 2.79 -1.70 -1.04 -0.67 0.30 91.2
21 GSW 1.75 3.64 -1.89 0.95 -2.84 0.61 94.1
22 1 LAC -2.03 0.79 -2.82 0.34 -3.16 0.58 91.7
23 1 DET 0.82 5.09 -4.27 0.13 -4.41 -0.27 88.6
24 -2 CHA -3.77 0.73 -4.49 -1.60 -2.89 -0.09 89.0
25 1 SAC -4.06 1.13 -5.19 0.59 -5.78 0.31 94.5
26 1 MIN -2.97 3.44 -6.41 -0.25 -6.16 0.49 96.0
27 1 NJN -3.84 2.87 -6.71 -0.14 -6.57 0.09 89.3
28 -3 TOR -0.97 5.76 -6.73 -0.95 -5.78 0.07 92.5
29 WAS -4.78 3.16 -7.94 -0.86 -7.08 -0.14 93.4
30 CLE -5.43 4.23 -9.66 0.85 -10.51 -0.06 92.8

That’s the basics, covered elsewhere. Next, let’s go Bayesian!

Rk Team Eff Dif. Bayesian Bayes OE Bayes DE Pace Max L8 L18 L62 Stdev
1 CHI 7.27 7.14 1.59 -4.71 89.3 7.42 7.86 7.86 12.23
2 LAL 6.65 5.39 2.06 -2.69 90.0 6.89 6.89 6.89 12.19
3 MIA 7.41 6.42 3.94 -1.72 90.0 6.42 6.42 7.86 12.98
4 DEN 4.96 5.18 3.64 -0.93 94.9 5.41 5.41 5.41 11.66
5 SAS 5.98 4.70 3.35 -0.79 91.3 4.92 5.28 6.95 12.25
6 ORL 5.36 4.76 0.46 -3.73 89.0 4.76 5.25 5.67 10.36
7 BOS 5.44 3.61 -1.43 -4.62 88.8 4.38 4.71 8.74 11.41
8 DAL 4.63 4.62 1.59 -2.48 91.8 4.62 4.62 5.75 11.82
9 OKC 3.67 4.32 3.59 -0.22 91.7 4.59 4.59 4.59 12.16
10 MEM 2.71 3.01 0.97 -1.69 90.2 3.68 3.68 3.68 12.00
11 POR 2.32 2.63 1.99 -0.33 86.6 3.33 3.36 3.36 13.21
12 HOU 2.33 2.72 3.15 0.75 93.0 3.01 3.18 3.18 11.83
13 PHI 1.34 1.08 -0.23 -1.18 90.4 2.75 2.75 3.03 12.03
14 NOH 1.11 -0.33 -0.46 -0.17 92.7 1.15 1.55 4.18 12.05
15 NYK 1.00 0.92 3.92 3.11 93.9 1.23 1.51 2.07 11.79
16 PHO -0.57 -0.75 1.55 2.21 93.5 -0.40 0.40 1.04 10.83
17 ATL -1.01 -2.34 -1.12 0.95 88.5 -0.63 0.12 2.77 13.35
18 GSW -1.89 -0.74 1.38 2.03 94.0 -0.74 -0.74 -0.74 12.92
19 MIL -1.39 -1.08 -4.47 -3.52 89.1 -0.87 -0.87 0.02 12.02
20 IND -1.40 -1.54 -1.56 -0.20 93.9 -1.27 -0.95 1.35 11.64
21 LAC -2.82 -2.37 -1.82 0.27 92.7 -2.26 -1.62 -0.44 10.71
22 UTA -1.70 -2.68 0.51 2.87 94.7 -2.68 -2.11 4.37 11.34
23 SAC -5.19 -3.64 -2.43 0.77 95.4 -3.17 -3.17 -3.17 12.56
24 DET -4.27 -3.27 1.67 4.55 89.8 -3.21 -3.21 -2.95 10.19
25 CHA -4.49 -4.25 -2.16 1.58 88.5 -3.96 -3.39 -1.76 12.39
26 MIN -6.41 -6.28 -2.79 2.74 95.0 -4.98 -3.89 -3.43 12.31
27 NJN -6.71 -6.28 -2.43 3.11 90.0 -5.73 -4.93 -3.65 9.88
28 TOR -6.73 -6.45 -0.58 5.11 91.7 -6.21 -5.87 -2.15 11.29
29 WAS -7.94 -6.89 -3.48 2.59 93.5 -6.73 -6.73 -4.49 12.93
30 CLE -9.66 -7.62 -3.29 3.42 91.8 -7.62 -7.62 -6.86 12.33

I sorted the Bayesian table by the maximum rating in the last 18 games, to try to avoid a few teams’ (LAL) tanking at the end. The 18 was chosen to allow 5 or 6 games since the trade deadline for the ratings to stabilize somewhat with the new players.

The standard deviation was calculated by taking the error for each of the last 30 games and running Stdev = Sqrt(average((Pred-Actual)^2)). Since this is a limited sample, the standard deviation for each team was regressed toward the average (12.0), using the prior being the overall distribution and the error for the standard deviation = 0.71*stdev/sqrt(N), where N was 30 for the number of observations.

Here is a Google Motion Chart showing how the Bayesian Ratings have progressed:

EDIT: Bayesian Power Rating Spreadsheet and Data here, for the first time ever!

Tags: , , , , , ,

One Response to NBA Final Regular Season Bayesian Ratings

  1. Stathead » Blog Archive on April 15, 2011 at 5:04 pm

    [...] the DStats today: Final regular-season Bayesian ratings, 2011 team efficiency charts, and a WOWY compilation. Posted on Friday, April 15th, 2011 at 6:03 [...]

Leave a Reply

Your email address will not be published. Required fields are marked *

Current day month ye@r *

DSMok1 on Twitter

To-Do List

  1. Salary and contract value discussions and charts
  2. Multi-year APM/RAPM with aging incorporated
  3. Revise ASPM based on multi-year RAPM with aging
  4. ASPM within-year stability/cross validation
  5. Historical ASPM Tableau visualizations
  6. Create Excel VBA recursive web scraping tutorial
  7. Comparison of residual exponents for rankings
  8. Comparison of various "value metrics" ability to "explain" wins
  9. Publication of spreadsheets used
  10. Work on using Bayesian priors in Adjusted +/-
  11. Work on K-Means clustering for player categorization
  12. Learn ridge regression
  13. Temporally locally-weighted rankings
  14. WOWY as validation of replacement level
  15. Revise ASPM with latest RAPM data
  16. Conversion of ASPM to" wins"
  17. Lineup Bayesian APM
  18. Lineup RAPM
  19. Learn SQL