However, a theoretical understanding of their success in practice is still a mystery. Moreover, recent advances [34] on fast convergence in games are limited to no-regret algorithms such as online mirror descent, which satisfy stability.

However, a theoretical understanding of their success in practice is still a mystery. Moreover, recent advances [34] on fast convergence in games are limited to no-regret algorithms such as online mirror descent, which satisfy stability.

Neural Information Processing Systems http://nips.cc/

Monte Carlo Counterfactual Regret Minimization (MCCFR) has emerged as a cornerstone algorithm for solving extensive-form games, but its integration with deep neural networks introduces scale-dependent challenges that manifest differently across game complexities. This paper presents a comprehensive analysis of how neural MCCFR component effectiveness varies with game scale and proposes an adaptive framework for selective component deployment. We identify that theoretical risks such as nonstationary target distribution shifts, action support collapse, variance explosion, and warm-starting bias have scale-dependent manifestation patterns, requiring different mitigation strategies for small versus large games. Our proposed Robust Deep MCCFR framework incorporates target networks with delayed updates, uniform exploration mixing, variance-aware training objectives, and comprehensive diagnostic monitoring. Through systematic ablation studies on Kuhn and Leduc Poker, we demonstrate scale-dependent component effectiveness and identify critical component interactions. The best configuration achieves final exploitability of 0.0628 on Kuhn Poker, representing a 60% improvement over the classical framework (0.156). On the more complex Leduc Poker domain, selective component usage achieves exploitability of 0.2386, a 23.5% improvement over the classical framework (0.3703) and highlighting the importance of careful component selection over comprehensive mitigation. Our contributions include: (1) a formal theoretical analysis of risks in neural MCCFR, (2) a principled mitigation framework with convergence guarantees, (3) comprehensive multi-scale experimental validation revealing scale-dependent component interactions, and (4) practical guidelines for deployment in larger games.

Sequential decision-making with multiple agents and imperfect information is commonly modeled as an extensive game. One efficient method for computing Nash equilibria in large, zero-sum, imperfect information games is counterfactual regret minimization (CFR). In the domain of poker, CFR has proven effective, particularly when using a domain-specific augmentation involving chance outcome sampling. In this paper, we describe a general family of domain-independent CFR sample-based algorithms called Monte Carlo counterfactual regret minimization (MCCFR) of which the original and poker-specific versions are special cases. We start by showing that MCCFR performs the same regret updates as CFR on expectation. Then, we introduce two sampling schemes: outcome sampling and external sampling, showing that both have bounded overall regret with high probability. Thus, they can compute an approximate equilibrium using self-play. Finally, we prove a new tighter bound on the regret for the original CFR algorithm and relate this new bound to MCCFR's bounds. We show empirically that, although the sample-based algorithms require more iterations, their lower cost per iteration can lead to dramatically faster convergence in various games.

This paper investigates how perturbation does and does not improve the Follow-the-Regularized-Leader (FTRL) algorithm in imperfect-information extensive-form games. Perturbing the expected payoffs guarantees that the FTRL dynamics reach an approximate equilibrium, and proper adjustments of the magnitude of the perturbation lead to a Nash equilibrium (\textit{last-iterate convergence}). This approach is robust even when payoffs are estimated using sampling -- as is the case for large games -- while the optimistic approach often becomes unstable. Building upon those insights, we first develop a general framework for perturbed FTRL algorithms under \textit{sampling}. We then empirically show that in the last-iterate sense, the perturbed FTRL consistently outperforms the non-perturbed FTRL. We further identify a divergence function that reduces the variance of the estimates for perturbed payoffs, with which it significantly outperforms the prior algorithms on Leduc poker (whose structure is more asymmetric in a sense than that of the other benchmark games) and consistently performs smooth convergence behavior on all the benchmark games.

Policy gradient methods have become a staple of any single-agent reinforcement learning toolbox, due to their combination of desirable properties: iterate convergence, efficient use of stochastic trajectory feedback, and theoretically-sound avoidance of importance sampling corrections. In multi-agent imperfect-information settings (extensive-form games), however, it is still unknown whether the same desiderata can be guaranteed while retaining theoretical guarantees. Instead, sound methods for extensive-form games rely on approximating counterfactual values (as opposed to Q values), which are incompatible with policy gradient methodologies. In this paper, we investigate whether policy gradient can be safely used in two-player zero-sum imperfect-information extensive-form games (EFGs). W e establish positive results, showing for the first time that a policy gradient method leads to provable best-iterate convergence to a regularized Nash equilibrium in self-play .