What-if Trade Machine

1 Same model as the season simulator. Each team starts from its 2025-26 schedule-adjusted rating; its current roster is valued player by player (box-score Game Score per minute above a replacement player, converted to points of team margin).

2 Minutes are re-shuffled. After the trade, each roster's projected minutes are handed out best players first until 240 minutes a game are filled, so adding a star pushes the weakest minutes off the floor rather than simply adding his value.

3 Discounted, as the backtest says. Only 52% of the roster change reaches the rating (fit by projecting 2025-26 from 2024-25): box scores overstate how much one move changes a team.

4 Re-play the season. Every remaining game is simulated with home court worth 1.7 points and a game-to-game spread of 12.5 points, plus a ±2.75-point uncertainty in each team's true strength; then the play-in and best-of-seven playoffs.

Not checked: salary-matching and cap rules, roster limits, trade restrictions, draft picks. We don't have contract data, so any combination of players is allowed. Rookies and players without a recent NBA season are valued at replacement level, and box scores miss most defense. Research estimates, not predictions of any single game.