Case under review

Precedent strength: 65/100 Evidence strength: 55/100 Jurisdiction: State

Case complexity: 45/100

Forest hyperparameters

Number of trees: 80 Max tree depth: 6 Random features / split: 2 of 4

Ensemble result

P(plaintiff wins)—

Votes plaintiff / defendant—

Tree agreement rate—

Out-of-bag accuracy—

Training set size—

Each cube in the grid is one decision tree, trained on its own bootstrap resample of the training set with a random subset of features considered at every split. Cube colour = that tree's live vote for the case on the left; brightness = the leaf's internal confidence. The bar below is the resulting ensemble probability.