How Our NBA Title Simulations Work — 10,000 Seasons and a Predicted Bracket

"Who wins tonight?" and "who wins the title?" sound like the same kind of question. They are not. The first is about one game between two teams whose rosters you already know. The second is about a five-month maze: 1,200 regular-season games still to come, a play-in scramble, four rounds of best-of-seven, and matchups nobody can name yet because they depend on results that haven't happened.

This article explains — in plain language — how our title simulations answer the second question, how we turn 10,000 simulated seasons into a single predicted playoff bracket, and why all of this requires a completely different machine than our daily game predictions.

Two Different Questions, Two Different Machines

Our daily NBA predictions are built by neural networks that study each scheduled game in detail: the confirmed starting five, the injury report, rest, recent form, and hundreds of other signals about the two specific teams playing that night. That depth is exactly what you want for "who wins tonight?" — and exactly what you can't have for a second-round series that may or may not happen next month, between two teams to be determined.

A title forecast has to price every possible matchup, including hypothetical ones. For that, we use our team strength ratings — the same Elo system behind our power rankings, rebuilt from nearly thirty seasons of NBA results. Ratings have a beautiful property: any two teams, at any moment, have a defined win probability against each other. That's the key that unlocks the whole bracket.

Then We Play the Season. 10,000 Times.

Here's the whole trick, honestly stated:

1. Lock in reality. Games already played are facts — their results stand, always. A playoff series our model sees at 3-1 is resumed from 3-1.
2. Flip weighted coins. Every game still to be played gets decided by chance, weighted by the teams' ratings — with home court and back-to-back fatigue folded in. If a team has a 65% chance to win, it wins 65% of the time and loses 35% of it, because upsets are part of basketball, not a bug in it.
3. Run the whole thing. Finish the regular season, set the standings, play the play-in, then every best-of-7 series under the real 2-2-1-1-1 home-court format — all the way to a champion. Write it down, reset, repeat. 10,000 times.

When the dust settles, the statistics are just counting. Won 1,200 of 10,000 simulated seasons? Your title chance is 12%. Reached the Conference Finals in 3,100 of them? That's a 31% chance of making the final four. The same 10,000 seasons answer every question at once — championship odds, the chance of reaching any round, and the full bracket below.

The Predicted Bracket — and "OKC in 6"

This is the part we're proudest of. Because every simulated season plays out a complete bracket, we can do more than rank teams — we can show the playoff bracket we expect, series by series. For each slot we surface:

• The most likely matchup. Before the field is set, a bracket slot can host different pairings — we show the one that came up most often, and how often.
• Who wins the series. Each team's probability of taking that series, conditional on the matchup happening.
• The most likely score. "OKC in 6 · 4-2 · 36%" means that in the simulations where the matchup happens, Oklahoma City closing it out 4-2 was the single most common outcome. Tap any series and you'll see the full distribution — 4-0, 4-1, 4-2, 4-3 — for both sides.

It's the difference between "this team is favored" and "here is the exact path we think the playoffs take, and how confident we are at every fork." When real series are underway, their actual score is locked in and the rest is re-simulated from there.

Why Not Just Use the Standings?

Because the path is half the story. Imagine two teams with identical records. One is the top seed in a soft half of its conference; the other has to get past two contenders just to reach the conference finals. Same record, very different title chances — and the simulation sees it automatically, because it walks both of their actual playoff paths thousands of times. The best regular-season record and the best title odds are often not the same team.

This is also why a team's odds can rise without it playing a game: when a dangerous rival is upset on the other side of the bracket, the simulations immediately reflect the easier road ahead.

What "30%" Actually Means

A 30% title chance means: run this season from today's situation many times, and that team lifts the trophy in about 3 of every 10. It does not mean "they won't win" — and it doesn't mean we were wrong if they don't. Shock champions happen, and an honest forecast gives them honest (small) numbers in advance, rather than pretending the favourite is a sure thing.

What makes a probability trustworthy is calibration: when we say 40%, it should happen about 40% of the time. You can inspect how we hold ourselves to that standard on the model transparency page.

Our Track Record, In Numbers

We tested the engine on 9 past NBA seasons (2016-2025), re-forecasting each one from several points of the season — with the model only knowing what was knowable then, trained on data through 2015. Across all 50 forecasts, the eventual champion was our single top pick 36% of the time and sat in our top 5 80% of the time:

For context: with 30 teams, naming the champion at random succeeds about 3% of the time. Forecasts are sharpest around midseason — enough games to know who's real, before late-season rest muddies the picture. The numbers above come straight from the backtest artifacts and update whenever the engine is retrained and re-tested — never by hand.

Forecasts That Sharpen As the Season Unfolds

A title forecast is alive. After every night of basketball, the picture changes: seeds firm up, the play-in field takes shape, and once the playoffs start, every series result narrows the bracket. On every prediction run we lock in the new results, refresh the ratings, and re-run all 10,000 seasons — so the numbers you see in May are conditioned on everything that's already happened, not on October's guesses.

The Fine Print (Kept Short)

• Home court and back-to-backs are priced in: a team on the second night of a back-to-back is rated a notch lower for that game, and the 2-2-1-1-1 series format gives the higher seed its edge.
• Reality always wins. Played games and decided series are facts, not guesses — the forecast only simulates what's still undecided.
• Randomness is controlled. 10,000 runs is enough that the numbers are stable to a few tenths of a percentage point; day-to-day changes you see are new information, not noise.
• What it doesn't know: a rating can't see a mid-playoff injury until it shows up in results, and a team that rests its stars in April looks worse than it is. We flag these honestly rather than pretend otherwise.

That's the system: ratings that always have an answer, a season played out 10,000 times, a bracket we'll commit to series by series, and probabilities you can hold us accountable for. The rest is basketball.