A video showing someone using real-time assistance (RTA) on ACR went viral this week. ACR says it identified and banned the account, which had played only at microstakes. The video raised a very current concern: if AI makes software easier to build, RTA will become easier to access, more people will cheat, and online poker will eventually become impossible to beat. Poker.org
I expect RTA to become more accessible. I don’t expect it to kill online poker. I’ve helped investigate cheating operations, including widespread cheating on Ignition. What makes me optimistic is simple: poker sites are not standing still. The same advances that make cheating tools easier to build can also make security much better at finding the people who use them.
Poker sites already use automated systems to review play and flag suspicious accounts, and AI should make those systems much more effective. PokerStars says it uses machine learning and automated monitoring to flag suspicious play. PokerStars ClubWPT Gold runs on technology from A5 Labs, which says it uses machine learning to detect bots, collusion, RTA, and unusual gameplay patterns. A5 Labs As these tools improve, security teams should be able to review more hands, compare more accounts, and catch patterns that would be hard to find by hand. The same technology that lowers the cost of building RTA also lowers the cost of building tools to catch it, and the poker sites have a lot of data to work with.
These sites also have a strong long-term financial incentive to stop cheating. If players think the games are rigged or overrun with bots, they stop playing. That means less traffic and less rake. Sites that want to stay in business have every reason to keep the games clean.
The economics of RTA
In ordinary six-max cash games, pure GTO has consistently produced win rates of a few to several bb/100 against strong human opponents. Meanwhile, the standard deviation in win rate in these games is usually around 80–100 bb/100. RTA is not a silver bullet. The swings an RTA user has to endure are still massive.
Cheating with RTA at low stakes is not particularly rewarding. At 5 bb/100, an RTA user playing 300,000 hands a year at NL100 would make about $15,000 in expected profit. If they play roughly 300 hands per hour, that’s $15 per hour. And it comes with huge variance, plus the risk of getting caught, banned, and having their funds confiscated. At microstakes, manual RTA is pretty clearly not a worthwhile endeavor.
At NL1K+, the same 5 bb/100 win rate over 300,000 hands is worth $150,000. That is where RTA becomes more dangerous. But high-stakes games also receive far more scrutiny, not just by security teams, but by the players themselves. The pools are small, the regulars mostly know who each other are, and suspicious accounts get discussed quickly. On GG, CoinPoker, and ACR, high-stakes players share hands, compare notes, and report accounts they think are cheating. I’ve personally been part of several of these investigations.
Where I’m more concerned
Exploitative bot farms are a bigger threat. They’ve been observed to win at much higher rates and play far more volume than a person using RTA by hand. They are also much harder to build, and once you run many accounts, they tend to leave patterns behind. These bots currently run rampant on Ignition, but not because they are uncatchable. Based on the investigations I’ve worked on, Ignition has either failed to take the problem seriously or has tolerated it to an extent that is hard to explain. Mobius Poker
I’m also more concerned about RTA in high-value ICM spots. GTO Wizard just released a solver that can build custom preflop ICM solutions in seconds, including custom stacks, payouts, bounties, and up to nine players. That makes it possible to get help on a handful of final-table or bubble decisions that can be worth thousands of dollars in tournament equity. Unlike cash-game RTA, the cheater does not need months of volume for the edge to matter. GTO Wizard
GTO LAB released a similar custom ICM solver, then removed it over concerns about game integrity. It’s not clear yet how much tools like this will be abused, but I think players should push back on offering instant custom ICM solves. At minimum, these products should have common sense guardrails, like a delay long enough to make real-time use impractical.
Pessimists sound smart, optimists make money
People were worried about solver technology killing online poker a decade ago. In 2016, posters on TwoPlusTwo were predicting that poker solvers would make online cash games unviable within a few years. That never happened. Since then, plenty of players have built entire careers, made serious money, and retired from the game. When people see a new threat, they tend to project it forward and underestimate how much the games and the industry will adapt. TwoPlusTwo
That is what I mean by “pessimists sound smart, optimists make money.” Pessimism often starts with a problem that is easy to see, then assumes it will keep getting worse while everyone else stands still.
I expect RTA to become easier to build and use. I also expect poker sites to become better at detecting it, though some will do a much better job than others. I see no current evidence that RTA is making online poker unbeatable. As cheating tools improve, so will the tools used to catch cheaters.