Compare ownership patterns after quality-guided continuations. Weights are frequencies among accepted samples under your policy and filters, not calibrated future probabilities.
Ready.
D(A,B) = ½ Σ |ownershipA − ownershipB|. One complete black/white point reversal contributes 1. Change the distance slider to regroup these samples instantly, without another search. Groups are exploratory; a slider cut does not establish distinct future worlds.
Ownership: +1 black, −1 white; 0 can reflect uncertainty or neutrality. These signed expectations are not separate black / white / neutral probabilities. Hover a point for its exact value. Stones show the current position.
Every row and column is one accepted continuation. Dark: similar ownership; orange: the largest distance among these samples (shown above). Samples are ordered by their current group.
Hover the matrix to compare two continuations.