Neural fictitious self-play

What changed

Combined off-policy deep reinforcement learning of approximate best responses with supervised reservoir-memory learning of average behavior, supporting end-to-end imperfect-information self-play.

Scientific scope

Combined off-policy deep reinforcement learning of approximate best responses with supervised reservoir-memory learning of average behavior, supporting end-to-end imperfect-information self-play. Extends fictitious play/FSP and inherits DQN/reservoir sampling. Approximate neural best responses and finite buffers do not inherit exact fictitious-play convergence automatically; source results concern specified Leduc/limit-hold'em settings, not all multiplayer games.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Extends fictitious play/FSP and inherits DQN/reservoir sampling. Approximate neural best responses and finite buffers do not inherit exact fictitious-play convergence automatically; source results concern specified Leduc/limit-hold'em settings, not all multiplayer games.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Deep Reinforcement Learning from Self-Play in Imperfect-Information Games

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction