Do Poker Bots Actually Win?

Short answer: some do, most do not, and the ones that do usually win less than their operators expected. The strategy is the easy part. Everything around it is where bots fail.

Where the edge is real

Against weak fields – micro and low-stakes cash games, soft club tables – a solver-approximate strategy plays tighter, more aggressive and more consistent poker than most opponents. It does not tilt, does not get bored and does not chase. Research bots have beaten strong human players heads-up and in six-player settings, which settled the question of whether it is possible in principle.

Why most bots lose anyway

• Rake. At micro stakes rake can exceed the edge of a good strategy. A bot that beats opponents by a small margin can still lose to the house.
• Abstraction errors. A bot plays a compressed strategy; where its tables do not fit the spot, it makes mistakes that sharp regulars exploit.
• Recognition failures. Misread stack, missed pop-up, wrong button: each costs a pot.
• Downtime. App updates break profiles; the bot cannot play until a fix ships.
• Table selection. Bots typically sit wherever there is a seat. Humans pick tables. This alone flips many bots from winning to losing.

The cost nobody budgets for

Getting caught. When an account is closed the balance usually goes with it, and often the linked accounts too. A bot that earns a modest amount over three months and then loses the bankroll has not won. Detection is discussed here and legality here.

What this means for our bot

We think our bot has an edge on soft low-stakes club tables. We do not promise that it does, we do not publish win rates we cannot audit, and we will not tell a beta tester to expect profit. Testers who report honest results, including losing ones, are the point of the beta.