Boston Celtics

A strike with no bid is treated as unpriced. The market win total is a straight line between the two strikes that straddle 50%.

• Our estimate starts from the roster. Each player's on-court net points per 48 minutes (from his most recent full season, pulled toward zero when the sample is small) is averaged by minutes across the roster list, then turned into wins at about 2.9 wins per point of net.
• That win total is then pushed through the real 2026-27 schedule 20,000 times, using a home win probability for every game, to get the range and the playoff and play-in odds.
• Ratings built from games alone (6,153 games before October 20) are kept as a comparison. They fade old games over the offseason, which pulls every team toward average.
• The market number is the wins where Kalshi's "wins or more" price crosses 50%, read off a straight line between strikes five wins apart.

• Possessions are an estimate from box totals, not tracked possessions.
• The season simulation uses each game's win probability from the ratings and nothing else: no rating uncertainty, injuries, trades after the cutoff or in-season rating updates, so the 10th-90th percentile band is narrower than a real season's spread.
• The 2026-27 schedule file lists 1200 games (80 per team on average) rather than 1230. Each team's missing games ([2] per team) are played against an average opponent at a neutral site with no rest terms.
• Projection method: (d) roster values. In the walk-forward study its hold-out MAE was 8.3 wins (n = 60 team-seasons) against 9.0 for last season's wins regressed toward 41. Typical error is still about 8 wins per team, so treat every number as a wide estimate.
• Roster values use each player's box-score plus-minus, which is not adjusted for teammates. Players with no box-score history count as average at 10 minutes a game.
• Ties in the seeding are broken at random; head-to-head and division rules are not applied.
• Every team has a usable market median.
• Market median uses the midpoint of the yes bid and ask for each strike and a straight line between the two strikes (5 wins apart) that straddle 0.5; a strike with no bid is treated as unpriced.
• Rosters are the DraftKings Best Ball player list for draft group 150564, kept only where the player is on his team's current ESPN roster (ESPN team rosters, 2026-10-19): released and departed players are out. Players DraftKings lists as Out are not in the engine's pool, so they have no projected minutes and are not in the top 8; each team's moves block names them.
• Projected minutes are the engine's per-player estimates from box scores through 2026-04-12; they are not scaled to the team's 240 minutes and do not model new teammates.
• Box scores start in 2021-22; a player with no box score on file is marked as a rookie or new, which can include veterans returning from abroad. Ages come from the ESPN player directory (2025-26 file) and are missing for players it does not list.
• This team has 2 games the schedule file does not list yet; they are treated as games against an average opponent.