Open source — tooling we maintain and contribute to.

PokerBot.com maintains a small set of open-source utilities used in our operator-side audit and research work. The pieces released publicly are those that benefit the broader poker AI community without compromising client confidentiality or our operational tooling stack.

Last updated · May 21, 2026·3 min read
01 · Approach

What we open-source and what we don't.

The line is operator-confidentiality-driven, not ideological. Two categories of work:

  • What gets released publicly: research utilities, anonymisation pipelines, hand-history format adapters, generic analytical infrastructure. Anything where the value comes from being widely available and where releasing it doesn't expose any operator's specific configuration.
  • What stays internal: the actual decision engines that ship in Managed Liquidity engagements, integrity-detection models tuned against specific client populations, platform-integration adapters tied to specific operator credentials. Anything where public release would compromise a client's deployment or expose detection methodology to the bots we're trying to detect.

The current open-source surface is small intentionally. We don't release for marketing; we release where the artifact has standalone utility for the broader community.

02 · Projects

What's actually publicly available.

  • Hand-history format adapters. Parsers for common private-club platform hand-history formats. Useful for anyone building operator-side analytics tooling. Standalone utility regardless of whether you use any commercial integrity technology.
  • Anonymisation pipelines. Utilities for stripping player-identifying data from hand-history corpora before analysis. We use these internally for our research aggregates; releasing them publicly helps the broader community do anonymised analysis correctly.
  • Behavioural statistic computation libraries. Reference implementations of standard poker statistics (VPIP, PFR, three-bet, c-bet, etc.) computed against arbitrary hand-history corpora. The library is generic; what makes detection work in practice is downstream of these primitives.
  • Andrew Kuznetsov's research code. Some of Andrew's individual research and algorithm work is published openly on poker-ai.org and adjacent academic repositories. Separate from PokerBot.com's commercial tooling but adjacent enough that operators sometimes find it useful.

Repositories live on the team GitHub organisation; specific URLs and contribution guidance distributed to engaged operators on request. We don't promote the open-source work heavily because it serves a specific niche audience rather than general operator engagement.

03 · Contributing

How to engage with the projects.

  • Bug reports and pull requests through standard GitHub workflow on each repository. Maintained by the engineering team; response window varies by project activity.
  • Operator-side feature requests that overlap with our technology work get folded into operational planning rather than into open-source roadmaps. If you're an operator using one of the utilities and you'd like a feature, the right channel is your engagement contact rather than a GitHub issue.
  • Academic collaboration on research-oriented projects (the analytical libraries, anonymisation tooling, Andrew's algorithm work) goes through the research channel on the contact page. Different routing from operator engagement.

Want repository access?

Repositories are publicly visible on GitHub. Operator-side feature discussions route through engagement contact; academic collaboration through the research channel.