Wednesday, January 11, 2012

NBA Team Rankings - January 11, 2012


Here are the NBA Betting Market (BTM) Power Rankings as of Wednesday, January 11, 2012.  I plan on updating these on a weekly basis, probably around Wednesday or Thursday (I'm trying to find the optimum time and day for updates that catches both the Wednesday and Friday games).


The general idea behind the rankings is to use the point spreads and over/unders for recent games (the past three weeks) to derive a power ranking, with more weight given to more recent games.  There is also an adjustment that accounts for actual game outcomes.  Refer to the Methodology page and my post at Advanced NFL Stats Community for more detail.

Glossary

Here is a glossary of terms used in the rankings table below (the structure of the table and the metrics shown draws significant inspiration from the Advanced NFL Stats Team Efficiency Rankings):

  • GPF - Stands for "Generic Points Favored".  It's what you would expect a team to be favored by against a league average opponent at a neutral site.
  • oGPF - Stands for "Offensive Generic Points Favored".  The component of a team's total GPF attributable to its ability to score points (per game).
  • dGPF - Stands for "Defensive Generic Points Favored".  The component of a team's total GPF attributable to its ability to prevent the other team from scoring points (per game).
  • GWP - Stands for "Generic Win Probability".  I converted the GPF into a win probability using the following formula: 1/(1 + exp(-GPF/6.5)).
  • oRANK - The team's oGPF ranking.
  • dRANK - The team's dGPF ranking.



The BTM Rankings:


RANK TEAM GPF oGPF dGPF GWP oRANK dRANK
1 Heat 8.0 7.5 0.5 0.77 1 15
2 Bulls 7.0 1.0 6.0 0.75 12 1
3 Thunder 6.0 6.0 0.0 0.71 3 17
4 Seventysixers 4.0 2.0 2.0 0.65 9 7
5 Trailblazers 4.0 2.0 2.0 0.64 8 8
6 Spurs 3.5 3.0 0.5 0.63 6 13
7 Magic 3.0 0.0 3.0 0.62 14 5
8 Nuggets 3.0 7.0 -4.0 0.62 2 29
9 Lakers 3.0 -0.5 3.5 0.61 17 3
10 Clippers 3.0 5.0 -2.0 0.60 5 21
11 Mavericks 2.5 1.0 1.5 0.60 13 10
12 Celtics 2.5 -1.5 4.0 0.59 20 2
13 Hawks 2.0 -1.0 3.0 0.58 18 4
14 Knicks 1.5 5.0 -3.5 0.56 4 26
15 Pacers 1.5 0.0 1.5 0.55 15 11
16 Grizzlies 0.0 -1.5 1.5 0.50 21 12
17 Suns -0.5 1.5 -2.0 0.48 10 22
18 Rockets -1.0 2.0 -3.0 0.46 7 25
19 Bucks -1.5 -3.5 2.0 0.44 24 9
20 Timberwolves -2.5 1.5 -4.0 0.41 11 28
21 Jazz -2.5 -2.0 -0.5 0.40 22 19
22 Warriors -3.0 -2.0 -1.0 0.39 23 20
23 Raptors -4.0 -4.0 0.0 0.35 27 16
24 Hornets -4.5 -7.5 3.0 0.33 30 6
25 Cavaliers -4.5 -1.5 -3.0 0.33 19 27
26 Kings -5.5 -0.5 -5.0 0.30 16 30
27 Pistons -5.5 -6.0 0.5 0.29 29 14
28 Bobcats -6.0 -3.5 -2.5 0.29 25 23
29 Nets -6.0 -5.5 -0.5 0.28 28 18
30 Wizards -6.5 -4.0 -2.5 0.27 26 24

Additional Key Factors:
  • League Average Points Scored (APS): 95.0 points
  • Home Court Advantage (HCA): 3.5 points

How to Use This Table to Build Point Spreads and Over/Unders

  • Point Spread (in favor of home team) = HCA + (GPFhome - GPFaway)
  • Over/Under = 2*APS + (oGPFhome + oGPFaway - dGPFhome - dGPFaway)

For example, the Knicks play the Thunder at Oklahoma City on Saturday.  According to the ranking table above, here is the predicted point spread and over/under:

  • Point Spread (in favor of OKC) = 8.0 points (= 3.5 + 6.0 - 1.5 )
  • Over/Under = 204.5 points ( = 2*95 + 6.0 + 5.0 - 0.0 - (-3.5) )

My goal here is transparency, so if there are any questions on these rankings or the methodology, please comment.

7 comments:

  1. Nicely done, Michael.

    One thing I'll note--when teams skew toward one side (all O or all D), it is perhaps a function of pace more than actual offensive or defensive efficiency.

    Evan Zamir is publishing similar rankings at http://thecity2.com/vegas-power-ratings/

    ReplyDelete
  2. Thanks DS. That's a good point on pace. There's a similar issue with the NFL rankings owing to the fact that scores are anti-correlated - a bad offense hurts its defense by giving the opponent favorable field position.

    At one point, I considered trying to divide out average number of possessions, but ultimately decided against it. I wanted to keep these rankings "pure" in that they ignore fundamental stats entirely, and focus solely on betting market metrics.

    Thanks for the link to Evan's site. I'll check it out. In the comments at Advanced NFL Stats Community, it was pointed out that Nutshell Sports has a vegas ranking as well (for NFL - not sure about other sports).

    ReplyDelete
  3. Great to see that you were able to do this in the NBA too, MB, in addition to your work in the NFL.

    I've been trying to eyeball market ratings at the apbr forums. DSMok has already posted a link to this page in that discussion. Will be great to have your more systematic approach in the mix.

    Interested to see what happens with your Over/Under projections. Have you thought about doing a simplified version of your current format for that? Meaning...a "generic total posted" for each team or something...list them from high to low based on market reads.

    I remember that back in the 80's we used to use simple algebra off the actual on-court team totals to make projections (Team A times Team B divided by the league average). Helped find some edges on the extremes...as run-and-gun teams would blow up into the 240's or higher when they squared off. Different era now though.

    Thanks again for tackling the NBA. Important contribution to the field I think...

    ReplyDelete
  4. Thanks Jeff. Maybe a Generic Over/Under (GOU)? The over/under you would expect for that team when playing a league average opponent? I think I could add that to next week's rankings.

    ReplyDelete
  5. The jazz should be higher they were on a 5 game streak until the lakers beat them.

    ReplyDelete
  6. I really like your blogpost.....

    ReplyDelete
  7. Very helpful suggestions that help in the optimizing topic,Thanks for your sharing.

    ซอมบี้

    ReplyDelete