Pick any five players, each as of one season between 2001–02 and 2025–26. The model adds up what each player brings, then applies combination effects: pairs of players who play differently together than apart. You get offense, defense and net (points scored minus points allowed), all per 100 possessions, with an 80% range for the lineup's long-run level (an 8-in-10 span). How the model works.
For players from other seasons: rated against their own season's league, or with their raw numbers as they were. Relative translates an older player's habits into today's league (a 2016–17 shooter takes even more threes now), so older lineups can come out several points apart between the two. A lineup of 2025–26 players gets the same forecast either way.
Loading the player list…
Loading the player list…
Keep four of Team A and try every player in the fifth spot, against a league-average lineup or Team B's five, to find who fits best. Each candidate gets the full forecast, combination effects and range. How the search works.
Pairings that play differently together than apart, beyond what adding up players assumes: learned from every pairing in real lineups, and 14 also measured one by one in real games. Each is small, but together they bend the forecast, most of all for lineups nobody would play. How they work.