You Are Here > Basketball-Reference.com > BBR Blog > NBA and College Basketball Analysis

SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all Basketball-Reference content, so you can quickly and easily find the content you want.

Also, our existing Basketball-Reference blog rss feed will be redirected to the new site's feed.

Basketball-Reference.com // Sports Reference

For more from Neil, check out his new work at BasketballProspectus.com.

BBR Rankings: 2009-11-20

Posted by Neil Paine on November 20, 2009

9450735_Heat_v_HawksOkay, it's time for the inaugural edition of the BBR Rankings, which will be coming to you every Friday from now until the end of the season. The idea behind these rankings is to rate a team's past accomplishments rather than predict what they'll do in the future -- sort of a retrodictive ranking rather than a predictive one. The Simple Rating System is the best predictor of future wins and losses because it focuses purely on point differential, filtering out the whims of lucky or unlucky performance in close games (which has been shown to have no bearing on future performance regardless of coach or players). However, it doesn't really give teams credit for wins and losses, which I feel like is the only real factor when you're looking at how impressive a team's season has been in the past.

To take a college football example, Oklahoma has a very impressive point differential this year against a very tough strength of schedule, and consequently their SRS is 5th-best in the country. Unfortunately for the Sooners, though, they also have lost four games. Would the Sooners likely win against most of the teams with fewer losses than they have? Yes, probably -- according to SRS they should be favored vs. everyone in the FBS except Texas, Florida, Alabama, and TCU. But no one in their right mind is going to rank OU ahead of undefeated Cincinnati and Boise State, or 1-loss Georgia Tech if we're talking about accomplishments, because the bottom line is that Oklahoma lost four games, and no amount of predictive advantage in the future is going to change that fact in the past.

All of this is to say that the premise of these rankings is to rate NBA teams on their past accomplishments rather than their likelihood of winning in the future. They're meant as a compliment to the SRS, and in fact, the basic structure of the ratings is the same as the SRS. Here's how they work: I calculated the average margin of victory for a winning team this year at home and on the road, and then averaged those together to determine the average point differential of a winning team on a neutral court (I didn't take a simple average of all winning teams because home teams are more likely to win by bigger margins, therefore skewing the value away from neutral). This is the baseline value that I give a team for winning (or deduct from them for losing). However, I also calculated the average MOV for teams at home this season, which is our HCA value. For teams that won at home, I gave them the avg. winner's MOV minus the HCA; for road winners, I gave them the winner's MOV plus the HCA; home losers received the avg. loser's MOV minus the HCA; and road losers got the loser's MOV plus the HCA. These were our static values for each team, crediting them only for wins and losses (and where each game took place). I then calculated the SRS normally (with HCA factored in), and those values were used for the SOS component of the BBR rankings. So, in effect, point differential does not matter when assessing the team's won-loss performance, to credit teams for "just winning" regardless of margin, but it does matter when assessing the strength of a team's opposition. I like this because there's no way to game the system by running up the score -- when you run up the score, you're actually only helping your opponents.

The end result is a ranking that takes into account strength of schedule in a legitimate, predictive way, but also credits a team for their own performance in the past. So, if you've managed to make it this far, here are the first BBR Rankings of the season:

Rank Team W L % SOS Rk
1 Atlanta Hawks 10 2 0.833 5
2 Denver Nuggets 8 3 0.727 15
3 Phoenix Suns 10 3 0.769 24
4 Los Angeles Lakers 9 3 0.750 2
5 Dallas Mavericks 9 3 0.750 19
6 Portland Trail Blazers 9 4 0.692 16
7 Houston Rockets 7 5 0.583 8
8 Orlando Magic 9 3 0.750 28
9 Boston Celtics 9 3 0.750 29
10 Chicago Bulls 6 5 0.545 6
11 Cleveland Cavaliers 8 4 0.667 26
12 Miami Heat 7 4 0.636 17
13 Oklahoma City Thunder 6 6 0.500 7
14 Utah Jazz 6 6 0.500 13
15 Sacramento Kings 5 5 0.500 10
16 Toronto Raptors 5 7 0.417 11
17 Detroit Pistons 5 7 0.417 12
18 Milwaukee Bucks 6 3 0.667 30
19 San Antonio Spurs 4 6 0.400 1
20 New Orleans Hornets 5 8 0.385 3
21 Indiana Pacers 5 4 0.556 22
22 Philadelphia 76ers 5 6 0.455 25
23 Washington Wizards 3 7 0.300 4
24 Charlotte Bobcats 3 8 0.273 14
25 Memphis Grizzlies 3 8 0.273 18
26 Los Angeles Clippers 4 9 0.308 23
27 Golden State Warriors 3 8 0.273 27
28 New York Knickerbockers 2 9 0.182 20
29 Minnesota Timberwolves 1 11 0.083 9
30 New Jersey Nets 0 12 0.000 21

ShareThis

9 Responses to “BBR Rankings: 2009-11-20”

  1. DSMok1 Says:

    Good job with the methodology! I like that formula a lot.

  2. P Middy Says:

    "All of this is to say that the premise of these rankings is to rate NBA teams on their past accomplishments rather than their likelihood of winning in the future."

    LIKES

  3. 123 Says:

    -since when is SRS a better predictor than point differential?

    -what's the point in creating a rating system that's not good in predicting future wins? If I want to know how good a team was in the past I'll just look at their W-L record

  4. Neil Paine Says:

    Since, like, forever. Remember, SRS is point differential, but it also adjusts for strength of schedule, so it can't help but be a better predictor than raw point diff.

    Secondly, the point is to credit teams for wins and losses in the past, even if those outcomes were lucky/unlucky. The advantage over straight W-L is that this method accounts for the location of the game, as well as how good/bad the teams they played were. Raw W-L treats all games the same, home or road, good or bad opponents.

  5. khandor Says:

    Neil,

    re: "The Simple Rating System is the best predictor of future wins and losses because it focuses purely on point differential, filtering out the whims of lucky or unlucky performance in close games (which has been shown to have no bearing on future performance regardless of coach or players)."

    Upon what basis would you make this specific claim, concerning the veracity of the SRS, as "the best predictor of future wins and losses"?

  6. Neil Paine Says:

    There's a great deal of research out there, but this springs to mind:

    http://sonicscentral.com/apbrmetrics/viewtopic.php?t=2060

  7. Jason J Says:

    I like that Josh Smith stopped shooting threes and started attacking the rim and immediately the Hawks jumped up a notch.

    I like that Melo started paying a little attention to defending and lost a touch of weight, and the Nuggets jumped up a notch.

    I like that Grant Hill decided to play like he was about 6 years younger than his actual age, and Jason Richardson started hitting the boards, and the Suns jumped up 3 notches.

    Nice system, Neil.

  8. khandor Says:

    Neil,

    LOL, :-), in a good way.

    By "the best predictor of 'future' wins and losses" you are NOT talking about a supposedly accurate methodology for deciding which of two teams involved in a specific contest which is still to be played in the future, is more likely to win this game than the other.

    What you really mean to say, instead, is that:

    "The SRS is the best 'predictor' of the specific number of total games a certain team is likely going to win during the following/next [i.e. subsequent] NBA season."

    In which case, I completely understand the veracity of THAT highly specific claim.

    Thanks for the clarification [in a "Back To The Future" sense].

  9. khandor Says:

    Site Operators,

    Kudos [and thanks] for taking care of that problem with "Ryan".