AI Versus Rule-Based Commercial Poker Bots: A 26,395-Hand Bot-Versus-Bot Comparison
Description
We evaluated six commercial poker bots purchased at the end of 2025. Vendors marketed these tools with claims of AI, GTO, or machine learning capabilities, yet published no long-term results. We ran these products in seven configurations against each other in a private PPPoker club. All accounts were controlled by the study, play used club chips with no cash value, and no human players participated. The study period ran from 27 November to 15 December 2025, covering NL4 6-max and 9-max games. We collected 26,395 hands in the main period. We excluded hands involving 3UpGaming from the primary analysis because it lost chips randomly to all opponents. This yielded a clean sample of 17,295 hands. We calculated 95% confidence intervals using percentile bootstrap over 30 table sessions with 2,000 resamples. The two machine-learning bots, which used vendor-side decision engines, formed one group. Four rule-based, profile-driven configurations formed the other. The AI bot group earned +29.2 bb/100 [+16.9, +43.1]. The rule-based group recorded −24.3 bb/100 [−33.5, −16.7]. These intervals do not overlap each other or zero. The sign of both group results held in every slice examined, including table size and date. A paid top profile did not improve rule-based performance, yielding −28.0 bb/100. The difference between the two AI bots remained smaller than the margin of error. Two of the four rule-based accounts, running on the Windows desktop client, were banned platform-wide by the anti-cheat system within the first four days; the AI bots, running in Android emulators, were not banned. A client update in mid-December 2025 ended the experiment by breaking all rule-based bots. These results are bot-versus-bot and must not be extrapolated to human play. We release the full hand history archive.
The record includes the full hand history archive (poker-bots-competition-hands.zip, SHA-256 9ea57179bf66e65aaf16427d54111e616290934a1a9ec5075728e74e775692d8).
Notes
Files
Kuznetsov_2026_AI_vs_rule-based_commercial_poker_bots.pdf
Additional details
Related works
- Is variant form of
- Other: https://poker-ai.org/ai-poker-bots-vs-rule-based-bots/ (URL)
Dates
- Collected
-
2025-11-27/2025-12-15Hand histories recorded on PPPoker
References
- Bowling, M., Burch, N., Johanson, M., Tammelin, O. (2015). Heads-up limit hold'em poker is solved. Science, 347(6218), 145–149. https://doi.org/10.1126/science.1259433
- Moravčík, M., Schmid, M., Burch, N., Lisý, V., Morrill, D., Bard, N., Davis, T., Waugh, K., Johanson, M., Bowling, M. (2017). DeepStack: Expert-level artificial intelligence in heads-up no-limit poker. Science, 356(6337), 508–513. https://doi.org/10.1126/science.aam6960
- Brown, N., Sandholm, T. (2018). Superhuman AI for heads-up no-limit poker: Libratus beats top professionals. Science, 359(6374), 418–424. https://doi.org/10.1126/science.aao1733
- Brown, N., Sandholm, T. (2019). Superhuman AI for multiplayer poker. Science, 365(6456), 885–890. https://doi.org/10.1126/science.aay2400
- Efron, B., Tibshirani, R. J. (1993). An Introduction to the Bootstrap. Chapman & Hall/CRC.
- Cameron, A. C., Gelbach, J. B., Miller, D. L. (2008). Bootstrap-based improvements for inference with clustered errors. The Review of Economics and Statistics, 90(3), 414–427. https://doi.org/10.1162/rest.90.3.414