More sports
NBA power rankings by the numbers (Elo)
Every NBA team is ranked below by an Elo rating built only from game results. The gap between two ratings gives a win chance, with the home edge included.
| # | Team | Rating | vs average | Record | Point diff | Games left: opponents | Playoffs | Title |
|---|---|---|---|---|---|---|---|---|
| 1▲2 | San Antonio SpursWest · 4–2 | 1,658 | 71% | 4–2 | +43 | 1,500 | 96.6% | 17% |
| 2 | New York KnicksEast · 2–4 | 1,646 | 69% | 2–4 | −4 | 1,496 | 94.9% | 14% |
| 3▼2 | Oklahoma City ThunderWest · 4–2 | 1,628 | 67% | 4–2 | −3 | 1,494 | 92.9% | 10% |
| 4 | Detroit PistonsEast · 5–1 | 1,626 | 67% | 5–1 | +48 | 1,495 | 95.6% | 11% |
| 5▲2 | Houston RocketsWest · 6–1 | 1,612 | 65% | 6–1 | +87 | 1,501 | 91.6% | 8.7% |
| 6 | Boston CelticsEast · 4–1 | 1,599 | 63% | 4–1 | +46 | 1,499 | 90.2% | 6.9% |
| 7▼2 | Minnesota TimberwolvesWest · 4–3 | 1,587 | 62% | 4–3 | +79 | 1,505 | 82% | 5.2% |
| 8▲3 | Cleveland CavaliersEast · 4–2 | 1,572 | 60% | 4–2 | +30 | 1,502 | 84% | 3.9% |
| 9▼1 | Denver NuggetsWest · 3–2 | 1,563 | 58% | 3–2 | −6 | 1,505 | 74% | 3.4% |
| 10▲3 | Philadelphia 76ersEast · 6–1 | 1,563 | 58% | 6–1 | +27 | 1,496 | 87% | 3.6% |
| 11▼2 | Los Angeles LakersWest · 4–2 | 1,562 | 58% | 4–2 | +31 | 1,508 | 75% | 3.2% |
| 12 | Toronto RaptorsEast · 3–2 | 1,548 | 56% | 3–2 | +53 | 1,506 | 73% | 2.9% |
| 13▲1 | Portland Trail BlazersWest · 4–1 | 1,538 | 55% | 4–1 | +57 | 1,513 | 65% | 1.9% |
| 14▼4 | Atlanta HawksEast · 3–4 | 1,527 | 53% | 3–4 | +6 | 1,497 | 67% | 1.5% |
| 15▲2 | Orlando MagicEast · 2–3 | 1,525 | 53% | 2–3 | −43 | 1,495 | 66% | 1.6% |
| 16▼1 | Charlotte HornetsEast · 3–4 | 1,512 | 51% | 3–4 | −57 | 1,507 | 56% | 1.3% |
| 17▲2 | Miami HeatEast · 2–4 | 1,498 | 49% | 2–4 | −19 | 1,499 | 50% | 0.8% |
| 18▲2 | Golden State WarriorsWest · 3–3 | 1,496 | 49% | 3–3 | +8 | 1,504 | 46% | 0.9% |
| 19▼3 | LA ClippersWest · 1–4 | 1,494 | 48% | 1–4 | −32 | 1,503 | 39% | 0.9% |
| 20▼2 | Phoenix SunsWest · 1–3 | 1,492 | 48% | 1–3 | −50 | 1,502 | 40% | 0.7% |
| 21▲1 | Dallas MavericksWest · 4–2 | 1,471 | 45% | 4–2 | +15 | 1,505 | 39% | 0.5% |
| 22▲3 | Memphis GrizzliesWest · 6–0 | 1,456 | 43% | 6–0 | +91 | 1,521 | 34% | 0.3% |
| 23▼2 | New Orleans PelicansWest · 2–4 | 1,450 | 42% | 2–4 | +11 | 1,513 | 20% | 0.2% |
| 24 | Indiana PacersEast · 2–4 | 1,394 | 35% | 2–4 | −19 | 1,511 | 12% | <0.1% |
| 25▼2 | Chicago BullsEast · 1–5 | 1,382 | 33% | 1–5 | −95 | 1,507 | 8.6% | <0.1% |
| 26▲1 | Milwaukee BucksEast · 1–5 | 1,369 | 31% | 1–5 | −79 | 1,510 | 5.5% | 0% |
| 27▲3 | Washington WizardsEast · 3–3 | 1,365 | 31% | 3–3 | +5 | 1,510 | 8.1% | <0.1% |
| 28▼2 | Sacramento KingsWest · 0–6 | 1,349 | 29% | 0–6 | −114 | 1,516 | 1.9% | <0.1% |
| 29▼1 | Utah JazzWest · 1–5 | 1,342 | 28% | 1–5 | −61 | 1,517 | 1.8% | <0.1% |
| 30▼1 | Brooklyn NetsEast · 0–5 | 1,326 | 26% | 0–5 | −55 | 1,512 | 2.0% | <0.1% |
Rating: Elo points; the league average is 1,505 right now and a 60-point edge makes the better team a 59% favourite on a neutral floor. “Games left: opponents” is the average rating of the teams still to play.
Head to head
Ratings: Raptors 1,548, Spurs 1,658. The home team gets 60 rating points.
In a best-of-seven: the Raptors win it 22% with home court, 18% without.
Updated 2026-11-01 19:25 UTC
How the chances are calculated
Ratings. Every team has an Elo rating built from every game since 1984–85 (FiveThirtyEight’s method; our code reproduces their published ratings game by game). Before a game the home team’s chance is 1 ÷ (1 + 10−(home rating − visitor rating + 60) ÷ 400): playing at home is worth 60 rating points, so between equal teams the home side wins 58.6%. After each game both ratings move by 20 × (result − chance) × a margin factor: a blowout moves them more than a one-point game, a favourite’s blowout less than an underdog’s. Each new season starts 40% of the way from last season’s rating to the average.
Settings. The step (20), the home edge (60) and the carry-over are the best of a grid on the seasons 2010–11 to 2022–23 (lowest log-loss of the pre-game chances); 2023–24 to 2025–26 were kept aside to test them.
Simulation. The games left are played 10,000 times with the ratings (a simulated result moves them too), the NBA tie-break procedure sets the seeds, then the Play-In and every best-of-seven (home court to the better record, games 1, 2, 5 and 7 at home) are played out.
Simplified. Simulated games have no score, so the last tie-break step (point differential) uses the games played so far; games abroad count as home games of the listed home team.
Fair odds are 1 ÷ chance: the price with no bookmaker margin. A rating knows only past results – not injuries, rest days or who sits out.
How good is the model
Hold-out test. The settings were fixed on the seasons 2010–11 to 2022–23; then the ratings forecast every one of the 3,693 regular-season games of 2023–24 to 2025–26 before tip-off. Lower log-loss and Brier scores are better; “always the home team” gives the home side its win rate of the fit seasons, 57.9%.
| Forecast | Winner picked | Log-loss | Brier |
|---|---|---|---|
| Our ratings (Elo) | 66.9% | 0.607 | 0.210 |
| Always the home team | 54.7% | 0.691 | 0.249 |
| Coin flip (50/50) | 50.0% | 0.693 | 0.250 |
No market comparison. We found no free source of NBA closing betting lines that may be published, so this card cannot show how the ratings do against the betting market – which knows injuries and rest days and is usually sharper. Use our chances as a second opinion, not as a reason to bet.
This season, for real: over its 88 regular-season games so far the ratings picked the winner in 58% (home team 52%); log-loss 0.640 (coin 0.693).
Calibration on the hold-out games: when the ratings gave the home team a chance in a range, how often it won.
| We gave the home team | Games | Home team won |
|---|---|---|
| 0%–20% (on average 16%) | 90 | 7% |
| 20%–35% (on average 29%) | 442 | 25% |
| 35%–50% (on average 43%) | 744 | 40% |
| 50%–65% (on average 58%) | 1,031 | 55% |
| 65%–80% (on average 72%) | 924 | 69% |
| 80%–100% (on average 85%) | 462 | 85% |
In-sample, seasons 2010–11 to 2022–23 (15,421 games): ratings 66.0% winners picked, log-loss 0.613; home team 57.9%.
NBA standings and playoff odds →NBA play-in tournament: format and odds →Raptors schedule 2026-27 with win chances →
Data: schedule and scores from balldontlie, checked every 6 hours against Wikipedia (team season pages and standings; licence CC BY-SA) – a score or a record that differs stops the update. Rating history before 2015–16: FiveThirtyEight NBA Elo (CC BY). Rules: NBA tie-break procedures, NBA: Play-In Tournament.
How an NBA Elo rating moves
- Start from a ratingEvery team carries one rating number, built from all its past games.
- Winner takes pointsAfter each game, the winner gains rating points and the loser gives up the same amount.
- Bigger wins move moreA bigger margin moves more points. A favourite's rout counts less than an underdog's big win.
- Home edgeThe home team gets an edge in the win chance. At a neutral site, neither team gets it.
- Drift between seasonsBefore a new season, every rating is pulled part of the way back toward the league average.
Opinion rankings vs Elo rankings
Opinion power rankings
- Opinion rankings are usually set by people, often a group of writers
- Can weigh injuries, trades and recent form
- The reasons can change from week to week
- No win chance attached to the order
Elo power rankings
- One rule applied to every team
- Uses results only: who won, by how much, where
- The rating gap turns into a win chance
Rating gap and win chance
| Rating gap | Win chance | What it means |
|---|---|---|
| No gap, neutral site | Even | Like a coin flip |
| No gap, home game | Home team slightly ahead | Home edge only: the home team is a slight favourite |
| Small gap | A bit above even | Close game, either side wins often |
| Medium gap | Clear favourite | The lower team still wins many games |
| Large gap | Strong favourite | Upsets still happen |
The tool above shows the exact chance for any two teams. Turn a chance into a price with the odds converter.
Where the numbers come from
Long history
Past ratings come from FiveThirtyEight's public NBA Elo file, licence CC BY 4.0. Our implementation reproduces their published ratings.
New games
Scores come from the balldontlie API and are checked against Wikipedia's season pages (CC BY-SA). If a score or record does not match, nothing new is published.
Automatic refresh
Data refreshes by itself every few hours. The tool shows when it was updated and hides the numbers if the data is too old.
Accuracy check
Settings were chosen on older seasons and tested on recent ones they did not see, against a coin flip, always the home team, and forecast against result.
No market comparison
There is no free betting-market data with a licence, so the ratings are not compared with betting odds and are not shown to beat them.
What the rating misses
Injuries, rest days, trades and lineups are not in the model. A missing star changes nothing until the result is in.
Same method for football
The NFL power rankings use the same Elo method for every NFL team.
Sources
- github.com /fivethirtyeight/data · checked Nov 1, 2026
- en.wikipedia.org /wiki/2026–27_NBA_season · checked Nov 1, 2026
- balldontlie.io /terms · checked Nov 1, 2026
Questions
How are NBA power rankings calculated here?
Each team has an Elo rating built from game results only. The winner takes points from the loser, bigger wins and upsets move more, and ratings drift toward average between seasons.
Is the NBA Elo ranking the same as the standings?
No. Standings count wins and losses. Elo also weighs the opponent, the margin and home court. For the season table, see NBA standings and playoff odds.
How well does NBA Elo call past games?
The accuracy card above replays recent seasons the settings did not see and compares the ratings with a coin flip and with always picking the home team. It does not compare them with betting odds.
Do Elo ratings reset every NBA season?
No. Each rating is pulled part of the way back toward the league average, then moves again with every game. Each team plays 82 regular-season games.
Is an Elo win chance a betting tip?
A rating summarizes past results and is not a betting tip. It misses injuries and rest, and bets carry a margin. To see how chances relate to prices, read moneyline explained.