Deep CFR: learned counterfactual advantage and average strategy

What changed

Used sampled game-tree traversals and neural counterfactual-advantage estimation to approximate linear CFR without a manually specified state abstraction, with a separately learned average strategy.

Scientific scope

Used sampled game-tree traversals and neural counterfactual-advantage estimation to approximate linear CFR without a manually specified state abstraction, with a separately learned average strategy. Originalv1 idealized theorem uses infinite buffers, infinitely many sampled traversals and minimum possible approximation error; finite practical neural training is not thereby certified equilibrium convergence. Function approximation is a learned generalization, not absence of representation assumptions. Linear CFR and external sampling are inherited.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Originalv1 idealized theorem uses infinite buffers, infinitely many sampled traversals and minimum possible approximation error; finite practical neural training is not thereby certified equilibrium convergence. Function approximation is a learned generalization, not absence of representation assumptions. Linear CFR and external sampling are inherited.

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 Counterfactual Regret Minimization

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