Curious about the current state of poker bots in 2025—especially for Texas Hold’em cash games (HU/6-max).
- What architectures/approaches are strong today (e.g., CFR/MCCFR, depth-limited solving, GTO vs exploitative, RL with abstractions)?
- Any public benchmarks, ladders, or eval suites people trust?
- Practical experiences running or testing against bots (e.g., Slumbot)? Any reproducible open-source projects you recommend?
- How do folks think about ethics and platform policies around botting in real-money environments?
- If you’ve built one: which libraries, hardware, and datasets were most useful?
energy_floww•1h ago
- What architectures/approaches are strong today (e.g., CFR/MCCFR, depth-limited solving, GTO vs exploitative, RL with abstractions)? - Any public benchmarks, ladders, or eval suites people trust? - Practical experiences running or testing against bots (e.g., Slumbot)? Any reproducible open-source projects you recommend? - How do folks think about ethics and platform policies around botting in real-money environments? - If you’ve built one: which libraries, hardware, and datasets were most useful?
Example references: - DecisionHoldem (blueprint + safe depth-limited solving): https://github.com/AI-Decision/DecisionHoldem - Slumbot: https://www.slumbot.com/ - Paper (DecisionHoldem): https://arxiv.org/abs/2201.11580 - PokerBotAI (bot competition context): https://pokerbotai.com
Would love to hear real-world lessons, pitfalls, and what “state-of-the-art” looks like now.