Talisman Red's NBA Computer Ratings
Presented here are computer ratings and rankings for NBA and, hopefully, a system that can come somewhat close to predicting final game scores. The Predictions tab includes games scheduled for the next few days.
To calculate a hypothetical margin of victory for any other matchup, just use the "Points" column. The home advantage fluctuates based on games played, but is consistently about 2.5 to 3 points.
For the last couple of years, all overtime games are recorded as "ties" because this helps the score predictions be slightly more accurate. 20 Jan 2026 update: The performance this year has been, well, woof. I noticed an error that was keeping way too many old games in the predictions file. Hopefully things will improve from here forward.
If you need original text files for anything below, check the Description tab. Let me know if you see any issues. Thanks and glad you're here!
Everything is now in one scrollable page, instead of the tabs which required Javascript. Easier for phone users, I hope. Scrapers should not be impacted. Original text files are linked in the Description section.
Ratings
NBA Computer Ratings Last updated 25-Sep-2026 09 PM ET Rank Team W L Rating Points BCS 1 Oklahoma City 0 0 3.57 61.06 1 2 San Antonio 0 0 3.06 59.25 2 3 Detroit 0 0 2.48 57.06 3 4 New York 0 0 2.42 57.73 4 5 Boston 0 0 2.02 56.52 6 6 Houston 0 0 1.74 54.72 5 7 Denver 0 0 1.69 54.79 7 8 Cleveland 0 0 1.48 53.75 8 9 Minnesota 0 0 1.11 52.84 10 10 LA Lakers 0 0 1.07 51.70 9 Rank Team W L Rating Points BCS 11 Charlotte 0 0 0.80 54.05 17 12 Toronto 0 0 0.78 52.67 11 13 Atlanta 0 0 0.53 51.47 13 14 Miami 0 0 0.53 52.37 16 15 Orlando 0 0 0.51 51.13 12 16 Phoenix 0 0 0.50 51.46 15 17 Philadelphia 0 0 0.17 49.43 14 18 Portland 0 0 0.08 49.75 18 19 LA Clippers 0 0 0.01 50.67 19 20 Golden State 0 0 -0.33 49.68 20 Rank Team W L Rating Points BCS 21 Chicago 0 0 -1.60 44.90 22 22 New Orleans 0 0 -1.66 45.92 23 23 Milwaukee 0 0 -1.67 44.10 21 24 Dallas 0 0 -1.81 45.13 24 25 Memphis 0 0 -2.22 43.97 25 26 Utah 0 0 -2.78 41.85 26 27 Indiana 0 0 -2.88 42.39 29 28 Brooklyn 0 0 -2.98 40.79 28 29 Sacramento 0 0 -3.04 40.35 27 30 Washington 0 0 -3.59 38.50 30
Ratings by Division
Teams are sorted by computer ranking (in parentheses), not necessarily by win-loss percentage. As on the front page, overtime games get recorded as ties to make the predictions hopefully more accurate. Playoff games between division teams end up getting counted here, too.
NBA Computer Ratings (Last updated 25-Sep-2026 09 PM ET) --- Atlantic CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 New York 0 0 0 0 2.42 ( 4) 57.73 0 0 -99.00 ( 20) 2 Boston 0 0 0 0 2.02 ( 5) 56.52 0 0 -99.00 ( 2) 3 Toronto 0 0 0 0 0.78 ( 12) 52.67 0 0 -99.00 ( 28) 4 Philadelphia 0 0 0 0 0.17 ( 17) 49.43 0 0 -99.00 ( 23) 5 Brooklyn 0 0 0 0 -2.98 ( 28) 40.79 0 0 -99.00 ( 18) --- Central CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 Detroit 0 0 0 0 2.48 ( 3) 57.06 0 0 -99.00 ( 8) 2 Cleveland 0 0 0 0 1.48 ( 8) 53.75 0 0 -99.00 ( 5) 3 Chicago 0 0 0 0 -1.60 ( 21) 44.90 0 0 -99.00 ( 4) 4 Milwaukee 0 0 0 0 -1.67 ( 23) 44.10 0 0 -99.00 ( 16) 5 Indiana 0 0 0 0 -2.88 ( 27) 42.39 0 0 -99.00 ( 11) --- Southeast CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 Charlotte 0 0 0 0 0.80 ( 11) 54.05 0 0 -99.00 ( 3) 2 Atlanta 0 0 0 0 0.53 ( 13) 51.47 0 0 -99.00 ( 1) 3 Miami 0 0 0 0 0.53 ( 14) 52.37 0 0 -99.00 ( 15) 4 Orlando 0 0 0 0 0.51 ( 15) 51.13 0 0 -99.00 ( 22) 5 Washington 0 0 0 0 -3.59 ( 30) 38.50 0 0 -99.00 ( 30) --- Northwest CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 Oklahoma City 0 0 0 0 3.57 ( 1) 61.06 0 0 -99.00 ( 21) 2 Denver 0 0 0 0 1.69 ( 7) 54.79 0 0 -99.00 ( 7) 3 Minnesota 0 0 0 0 1.11 ( 9) 52.84 0 0 -99.00 ( 17) 4 Portland 0 0 0 0 0.08 ( 18) 49.75 0 0 -99.00 ( 25) 5 Utah 0 0 0 0 -2.78 ( 26) 41.85 0 0 -99.00 ( 29) --- Pacific CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 LA Lakers 0 0 0 0 1.07 ( 10) 51.70 0 0 -99.00 ( 13) 2 Phoenix 0 0 0 0 0.50 ( 16) 51.46 0 0 -99.00 ( 24) 3 LA Clippers 0 0 0 0 0.01 ( 19) 50.67 0 0 -99.00 ( 12) 4 Golden State 0 0 0 0 -0.33 ( 20) 49.68 0 0 -99.00 ( 9) 5 Sacramento 0 0 0 0 -3.04 ( 29) 40.35 0 0 -99.00 ( 26) --- Southeast CW CL W L Rating Rank Points NCWNCL Non-Conf Sched 1 San Antonio 0 0 0 0 3.06 ( 2) 59.25 0 0 -99.00 ( 27) 2 Houston 0 0 0 0 1.74 ( 6) 54.72 0 0 -99.00 ( 10) 3 New Orleans 0 0 0 0 -1.66 ( 22) 45.92 0 0 -99.00 ( 19) 4 Dallas 0 0 0 0 -1.81 ( 24) 45.13 0 0 -99.00 ( 6) 5 Memphis 0 0 0 0 -2.22 ( 25) 43.97 0 0 -99.00 ( 14)
Game Predictions
These predictions already include home advantage, if necessary.
PREDICTIONS FOR UPCOMING GAMES Date Away Team Home Team Total Prediction ----------- -------------------- -------------------- ----- --------------
Division Rankings
The average computer rating of all teams in a division is shown here.
Rank Conference Mean Rating ---- ------------------------- ----------- 1 Northwest 0.73 2 Atlantic 0.48 3 Southeast -0.18 4 Southeast -0.24 5 Pacific -0.36 6 Central -0.44
Description
Original Text Files
Feel free to download these; if you use them elsewhere, give credit to talismanred.com.
- Rankings
- Division Standings (but sorted by computer rating)
- Game Predictions
What is this about?
Computer ratings are the starting point for Vegas oddsmakers, and have traditionally done reasonably well at predicting game winners. On this page, I share my own formula, which I've been producing for one sport or another for about 20 years. If you're familiar with ratings you've probably heard of the others: Sagarin, Massey, and more.
So it's just a math formula?
Yep, that's all that computer ratings are. You can do it this way, or you can do like they do for college sports and do a human "poll" (coaches, writers, whoever).
The formula here is one that I developed quite a few years ago to try and "compete" with all the other folks who are doing it. That's where the predictions come from -- incorporating historical games and trying to use those to figure out what will happen next. It includes scores, dates, and locations, but no other information. No information about injuries, pace or speed of the game, shooting percentages, etc. Those things *do* matter, but they aren't factored in here.
Are the predictions any good?
In 2023-24, my NBA ratings predicted 64.2% of the correct game winners. By comparison, Jeff Sagarin's method got 65.8% correct. Over a season of about 850 games, that is a 10-15 game difference. I'll let you decide how much difference there is for 10-15 games out of 850. (I suspect that most of that was early in the season, when it's more difficult to know what a team's true strength is going to be. Jeff's preseason algorithm is probably more sophisticated than mine.)
That said, the Vegas line got 68.0% of winners. You just cannot beat a human touch!
In 2025-26, about mid-January I noticed that this year's predictions were doing terribly (61%, compared to most of the other computers at 65-66%). So I did what any self-respecting programmer would do, I adjusted the formula in hopes of improving accuracy. I'll report back with the results. One of the big factors is how much recent information do you keep when trying to predict a team's performance. A couple weeks? A month? The whole season?
Ratings Description
This rating system uses a formula developed by the owners of talismanred.com. It is explained in more detail over on the college football page, but the bottom line is that it can be used both for ranking teams as well as predicting future scores. In some circles, you'll see that referred to as both "retrodictive" and "predictive."
About the Author
I am a faculty member at Indiana University in Bloomington, with a Ph.D. in Atmospheric Science from the University of Alabama in Huntsville. For my day job, in addition to teaching lots of courses I have research interests in the broad areas of thunderstorms and numerical weather prediction (forecasting the weather using computers). Contact me using this email form if you have questions or non-hateful comments.