The league outlook
Ranked by projected mean wins within each conference: 1–6 top six, 7–10 play-in, 11–15 outside. These are projected places, not guaranteed finishes; the Top 6 column shows probabilities. Exact mean-win ties are ordered alphabetically. Select a team to explore. The range contains the middle 80% of forecast outcomes. Market 50% is an interpolated win-ladder crossing, not an expected win total.
| NBA # | Team / conference rank | Wins | 80% range | Since Oct 7 | Top 6 | Market 50% |
|---|
projected wins
How our view has changed
What's your call?
Make a prediction while you listen or talk it through. Saved only in this browser, never sent anywhere.
Model versus market · the same question
| Regular-season wins | Our chance | Market bid–ask | Gap to midpoint |
|---|
Distribution data
Probability of finishing with each exact win total. Rounded for display.
| Wins | Probability |
|---|
What went into this release?
We refreshed roster membership and injury evidence, carried forward the corrected historical inputs, and added a fitted player-impact adjustment. The new simulation also reconciles every season to exactly 1,230 league wins.
The impact adjustment improved historical win-distribution accuracy by about 3.5% in our development comparison. That supports trying it here; it is not proof of an advantage this season. The refreshed baseline is available above to keep the comparison visible.
What the model knows — and what is still uncertain
Player availability, projected minutes and offensive/defensive contributions feed team strength. The impact adjustment uses dated DARKO ratings where available and keeps the original player prior where they are missing. Its two coefficients are fitted on earlier seasons, with uncertainty carried into the simulations. We retained the registered calibration and lottery assumptions. No new generic bench adjustment or market blending was added.
Camp rosters remain provisional. Five retained long-absent players use the recent experimental returner prior; five new undrafted camp players use the existing undrafted-rookie prior. Provider injury return dates are uncertain assumptions, including dates labelled for reassessment or with no timetable. Day-to-day reports do not imply confirmed health. Preseason appearances help check roster evidence; preseason scoring is not an extra team-strength adjustment.
These are season distributions, not a game-by-game simulation of the released schedule. Conference probabilities rank the simulated win totals, with random ties rather than official tiebreakers. “Top 6” means a direct playoff place; “play-in places” means seeds 7–10, not winning the play-in. No championship odds are inferred.
The October 7 forecast and October 15 prospective study remain frozen. This is a separately dated forecast that can be scored later. Another refresh can incorporate the final roster decisions around October 19.
Sources, timing and market checks
Roster and injury inputs come from ESPN, with transaction conflicts checked against CBS's dated transaction ledger and the official Dallas coaching list. The roster deadline is listed on NBA key dates. Player-impact research uses DARKO.
Market comparisons use captured Kalshi regular-season “at least” win contracts. We show only active contracts with verified rules, positive displayed size on both sides, a two-sided spread no wider than 20 percentage points and a capture within 24 hours of the release cutoff. Nonmonotone or incomplete ladders do not get a 50% crossing. Gaps describe disagreement, not returns after fees or an instruction to trade. Quotes are dated and do not update here.