Results

From clueless to capable.

Every game the shared TrueUp has played with its HRA learner, which practised 8,000 games off screen before its first game here. Nobody told it the rules. The yellow points are evaluation games: boards it never trains on, played with learning and experimenting switched off. They're the honest measure of what it has learned.

Loading…

Games learned from

—

Lessons

—

one per move it learned from

Moves decided

—

every one a live TrueUp decision

Evaluation score

—

Score as it learns

Training gameAverage of the last 20Evaluation on unseen boards

How long it survives

Moves per game, averaged over the last 20. A game ends when TrueUp is caught or after 3,600 moves.

Caught vs. survived

—

Latest games

Game	Kind	Score	Level	Moves	Ended	Learned from