Thin gray line: per-update value (or bin mean); blue: rolling mean. Hover for values; "Table" opens the numbers.
Final pot of each of the learner's hands, in bb, blinds included and uncalled bets returned. Percentiles over each update's hands; each point averages them over the bin.
Rates over the learner's own decisions in each update's rollout, %. Thin gray: per update (or bin mean); blue: rolling mean. "—" means no opportunity in that update. Each panel has its own y-scale.
Last 40 non-JSON lines of the supervisor's log, newest last.