StaticExtractThe normalized leaderboard score of a model, read from its holdout metrics only and oriented so that higher is always better (a lower-is-better metric is negated, a closer-to-one metric is minus its distance from one), with the metric it was read from. The plan's success metric when the holdout gives a score on it; otherwise the model's primary proper score, the one holdout metric that is a proper score (a kept simple baseline stores that score alone, so it would otherwise never take a place on the leaderboard). Null when neither gives a score: the holdout holds no rankable value of the success metric and no single proper score. A missing score is never a stand-in number.
the model's HoldoutMetrics JSON
the plan's SuccessMetric (e.g. AUC, RMSE)
StaticIsWhether a score's metric is the plan's success metric (case- and whitespace-insensitive, as the score is read).
The leaderboard scores of an experiment's models.