Trade and lineup lab

Move players between teams and see what it does to each roster and its win total, or put any five players on the floor together and see how they project.

Impact only, not a legal trade checker. The lab ignores salary, the cap, aprons and roster size rules. It only asks how good each team would be.

The lab could not load right now. Try again in a minute.

How this works

The method

Roster rating. Each player gets his last season minutes per game, counted over at least 41 games so a short stint does not look like a starter, capped at 36. The team's 240 minutes go to players in that order. Each minute is worth the player's impact rating (RAPM) per 100 possessions, and minutes nobody covers go to a replacement level player (-2.0). It is the same roster rating the win probability model uses.

Trades. Moving a player changes both rosters, the minutes re-flow, and the new rosters are scored the same way. One point of roster rating has been worth 1.22 points of real net rating, so net change is roster change times 1.22. Wins change is net change times 2.14 wins per point, times the share of the 82 games still left.

Lineups. Offense is league average (116.6) plus the five offensive ratings times 1.22. A defensive rating counts points saved, so plus means good defense; defense is league average minus the five defensive ratings times 1.22. Net is offense minus defense.

What we measured

Wins per point: 2.14 wins over 82 games per point of net rating, a straight line through the 30 teams of last season. The usual rule of thumb is about 2.7; last season's win totals were more bunched than the margins, so the fitted line is flatter and trades move wins a bit less.
Roster to net rating: 1.22 points of net rating per point of roster rating, fitted for the win probability model.
Matches the win model: No roster ratings for this season yet.
Projected wins: Wins so far plus each remaining game's win probability, added up, plus the trade's change.

What it cannot see

Fit and chemistry: two ball-dominant stars do not simply add up, and each rating was earned in the player's old role.
Injuries, rest and age: minutes come from last season, not from who is healthy today.
One season of RAPM is noisy. Each player's 80% range (the bar in the lineup chart) shows how much; a trade that moves a team by less than a point is inside the noise.
Rookies and players with no rating count as replacement level (-2.0).
No salary, cap, apron or roster size rules. A trade the lab likes may be impossible in real life.