🏆 Scoreboard

Gains captured: how much of what was actually on this island got taken, with autarky at 0.0 and the frontier at 1.0. Not raw efficiency — two islands are not equally hard, and one where autarky already scores 0.82 has almost nothing on the table while one at 0.60 has a great deal, so ranking on raw efficiency ranks the draw. Below zero means the traders ended up worse off than if they had never traded, which is a real outcome and is not clamped.
The level is the format — traders, goods, days — and not the island: the island is drawn per round, so a seed is a roll rather than a level, while four traders face a different frontier from two and thirty days is more room to learn than three.
A game is one attempt, and may be more than one round: a one-round game scores that round, a several-round game scores the median of its rounds. The rounds have to be declared as one game before they are played — ten rounds played and the best three called a game afterwards is cherry-picking with a statistic on top.

Each trader against their own autarky, because raw Cobb-Douglas utilities are not comparable between traders. 1.00× is the line — below it, trade left them worse off than never trading. Ranked on the best round: a lucky island or a partner who happened to want what you could make still counts, which is what a high score is. The median and the round count sit beside it, so you can see whether a top score was one round or a habit.