Playoff odds · 2026-27 · 10,000 simulations · updated Sat, Oct 17
The Spurs lead the best-record race at 30%. The rest of the season, played out 10,000 times.
01Projected standings
Preseason: based on the 2026-27 schedule and preseason ratings, including summer roster moves. No games are final yet, so every team is 0-0 and the ranges are wide.
East
| # | Team | Seeds 1 to 15 | |||||
|---|---|---|---|---|---|---|---|
| 1 | Knicks | 0-0 | 64-18 | 99% | 3% | 90% | |
| 2 | Pistons | 0-0 | 56-26 | 93% | 17% | 63% | |
| 3 | Celtics | 0-0 | 53-29 | 87% | 28% | 47% | |
| 4 | Cavaliers | 0-0 | 52-30 | 85% | 31% | 44% | |
| 5 | Raptors | 0-0 | 51-31 | 83% | 33% | 39% | |
| 6 | Hornets | 0-0 | 49-33 | 76% | 41% | 31% | |
| 7 | Magic | 0-0 | 48-34 | 72% | 43% | 28% | |
| 8 | Heat | 0-0 | 47-35 | 69% | 47% | 24% | |
| 9 | 76ers | 0-0 | 45-37 | 62% | 51% | 18% | |
| 10 | Hawks | 0-0 | 44-38 | 58% | 54% | 15% | |
| 11 | Bucks | 0-0 | 26-56 | 5% | 16% | <1% | |
| 12 | Pacers | 0-0 | 25-57 | 5% | 13% | <1% | |
| 13 | Bulls | 0-0 | 25-57 | 5% | 14% | <1% | |
| 14 | Nets | 0-0 | 22-60 | 2% | 8% | <1% | |
| 15 | Wizards | 0-0 | 16-66 | <1% | 2% | 0% |
West
| # | Team | Seeds 1 to 15 | |||||
|---|---|---|---|---|---|---|---|
| 1 | Spurs | 0-0 | 65-17 | >99% | 1% | 95% | |
| 2 | Thunder | 0-0 | 64-18 | >99% | 1% | 94% | |
| 3 | Rockets | 0-0 | 52-30 | 91% | 18% | 56% | |
| 4 | Nuggets | 0-0 | 49-33 | 85% | 28% | 40% | |
| 5 | Suns | 0-0 | 46-36 | 78% | 34% | 31% | |
| 6 | Timberwolves | 0-0 | 45-37 | 77% | 37% | 27% | |
| 7 | Lakers | 0-0 | 44-38 | 70% | 41% | 22% | |
| 8 | Trail Blazers | 0-0 | 43-39 | 68% | 43% | 20% | |
| 9 | Clippers | 0-0 | 37-45 | 45% | 48% | 8% | |
| 10 | Warriors | 0-0 | 34-48 | 31% | 42% | 4% | |
| 11 | Pelicans | 0-0 | 30-52 | 21% | 35% | 2% | |
| 12 | Mavericks | 0-0 | 27-55 | 14% | 25% | <1% | |
| 13 | Kings | 0-0 | 25-57 | 8% | 18% | <1% | |
| 14 | Jazz | 0-0 | 23-59 | 6% | 15% | <1% | |
| 15 | Grizzlies | 0-0 | 23-59 | 6% | 13% | <1% |
Projected: average final record, with a bar from the 10th to the 90th percentile of wins (the dot is the average, the tick is 41 wins). Seeds: darker squares are more likely finishes, 1 on the left; gaps mark the play-in line after 6 and the cut after 10.
Best record in the league
04Odds over time
03Model vs markets
Market chance is the midpoint of the best bid and ask on Kalshi or Polymarket US (tap it to open the market). Win line: Polymarket's line when it has one, otherwise Kalshi's line closest to a coin flip; our chance of reaching it comes from the simulator's average and range. 15 teams are 20+ points apart from a market. Before games are played a big gap usually means our preseason ratings are missing something, like a star returning from injury, so read these as a check on the model, not as picks. Analysis, not advice. 21+.
How this works
The method
Every morning we play out the rest of the regular season 10,000 times. Games already played count as they finished. Each remaining game is decided by a weighted coin flip using the pregame win chance from our win probability model (team Elo ratings, home court and rest). A team with a 64% chance wins that game in about 64 of every 100 simulations. Games the NBA has not scheduled yet (the NBA Cup knockout nights are set in December) are filled in as neutral-court games against an average team so everyone plays 82.
A rating is an estimate, not the truth, so each simulation also nudges every team's rating up or down by a random amount (a normal draw with a spread of 80 Elo points, kept for that whole simulation). The base odds are stretched a little to make up for it, so on average each game still goes to the favorite as often as the model says. Without the nudge, the ranges came out too narrow in the test below. Then we rank each conference by wins. Seeds 7 to 10 play the play-in: 7 hosts 8 and the winner is the 7 seed, 9 hosts 10, and the loser of 7 vs 8 hosts the winner of 9 vs 10 for the 8 seed. Play-in games use the same Elo formula with the higher seed at home. Playoffs means a top-6 finish or winning a play-in spot. Top 4 means home court in the first round.
Magic numbers appear once games are played: own wins plus losses by the closest chasing team that clinch finishing ahead of it (82 + 1 minus own wins minus the chaser's losses). Play-in uses the best team outside the top 10 as the chaser, top 6 the best team outside the top 6. Ties are ignored, so a magic number can be one game conservative.
How we tested it
We ran the simulator on 2025-26 as if it were Thu, Jan 1 and Sun, Mar 1: only games before that date counted, and every later game used ratings frozen on that date (which is exactly what a live run does). Then we compared with how the season really ended. Wins miss is the average gap between projected and final wins. Brier is the average squared miss of a percentage (0 is perfect, lower is better).
| Method | Wins miss | Brier, playoffs | Brier, top 6 | Inside range |
|---|---|---|---|---|
| Simulator (this page) | 4.9 | 0.105 | 0.116 | 80% |
| Simulator without rating noise | 4.9 | 0.113 | 0.134 | 57% |
| Every game a coin flip | 7.9 | 0.093 | 0.089 | 23% |
| Current win % carried forward | 5.1 | 0.083 | 0.133 | n/a |
| Method | Wins miss | Brier, playoffs | Brier, top 6 | Inside range |
|---|---|---|---|---|
| Simulator (this page) | 1.9 | 0.098 | 0.056 | 80% |
| Simulator without rating noise | 1.9 | 0.108 | 0.051 | 77% |
| Every game a coin flip | 4.5 | 0.080 | 0.058 | 33% |
| Current win % carried forward | 2.6 | 0.083 | 0.067 | n/a |
What this shows: the projected wins beat carrying each team's current win % forward at both dates, and the ranges held the real final win total for about 80% of teams, as they should. The playoff odds did not beat the simple baselines in this one season. Only 30 teams are graded, a few surprises (Atlanta went 16-5 after March 1) swing the score, and the win model's coefficients were fit on that same season, a small head start. We will grade 2026-27 the same way when it ends.
What it cannot see
Injuries, trades and who is resting are invisible except through results. Ratings do not update inside a simulation, so a team that starts hot in one simulated season does not get stronger for it. Ties in the standings are broken by a random draw, not the NBA's head-to-head and division rules. Before opening night the odds rest only on last season's ratings, pulled back toward average, plus summer roster moves, which is where they are least sure.