Raul Moriarty

Background

As Communications Lead for Poker Bot AI (the umbrella organization that runs this website), I've worked on the edge of developing a new generation of software products; also, as a former entrepreneur/online poker player, I've been fortunate to work both in development of poker platforms and communication between developers and users. For the past 15 years, I've worked closely with engineers creating artificial intelligence to drive decision making; operators looking to protect their platform from cheating; and most importantly, end-users who want to know if what they're told about modern poker software is true. Those end-users are where we get our purpose.

Most public writing on "GGPoker bots," "hacks," and "cheating" is either a marketing surface claiming that solver anchored artificial intelligence software for use at the poker table is an exploit of the server or a forum thread treating any kind of automated decision support as the same thing as deck-prediction cheat snake oil. Both of these frameworks are incorrect and will prevent informed readers from understanding what real poker software does, where the engineering problems are, and what an operator like GGPoker actually is doing with its detection side.

Areas of focus

The threads I keep returning to in poker software:

Modern poker software architecture

The use of solver-anchored baselines (policy families developed by using CFR-based outputs from PioSolver, GTO+, MonkerSolver), compressed to enable querying during runtime, paired with opponent models which adjust and converge during an individual session as opposed to adjusting over years of HUD data.

The GGPoker ecosystem

Architectural features such as anonymous tables, rotating screen names, operator controlled HUDs, and use of PokerCraft analytics are some of the factors that have developed into the modern form of this problem. The NSUS Group has the largest amount of study as an operator within the network; therefore, most of the reader-submitted "practical" problems will be from a GGPoker perspective.

Detection from the operator side

The four-layer model (behavioral fingerprinting; statistical play-pattern analysis; anti-collusion graph models; human review) and where naive software is caught. This is the area of this subject that will benefit the most from honest explanation – not as a checklist of “how to avoid bans”, but as an adversarial-classification problem with an asymmetric cost matrix

Business and product

Fifteen years experience in developing software as an engineer provides me with a useful filter on how many of the claims made by Poker AI developers, have a basis in reality and how many are just pure Sales Copy. I am finding that most of what is referred to as a "Poker Hack", falls into the category of Sales Copy, and stating this as clearly as possible has been found to be much more helpful to my readers than providing yet another neutral review.

Game theory in practice

Where the math says "stop." Some spots in poker are solved well enough that further automation is rounding error; others — deep-stacked multiway turn play, ICM-heavy MTT endgames — are still meaningfully open. Knowing the difference is part of taking the field seriously.

About this site

Three long-form notes (hacks, detection, FAQ) plus the homepage cover what I think is worth saying publicly about the GGPoker ecosystem right now. Pages are revised when the field changes; dates at the top of each piece are the last revision, not the original publication.

We also include a chat-link for developers and curious players; if you have an issue with something that has been described as being implemented or your view of how things should work on the business side (or any other questions) we would be happy to hear from you. The poker bot AI team reads everything that is submitted.