The bot wars, from the operator's side

Every poker player has a bot story. The reg who never sleeps. The account with inhuman timing patterns. The forum thread with the damning stats, and the site that seemed to do nothing about it.

I’ve read those threads for twenty years. First as a player, nodding along. Then from inside two poker operators, where I learned the real shape of the problem. The bot question is a maturity test, and sites pass or fail it in public.

The maturity test

A sophisticated operator understands that banning bots and staying in business are the same project. The chain is short and it only runs one way. Fewer bots means a level playing field. A level playing field means a healthier ecosystem. A healthier ecosystem means happier players. Happier players means a business that lasts.

A less experienced site reads the same board backwards. Bots pay rake. Bots fill tables. Ban them and this quarter’s volume graph dips, so the temptation is to look away and book the revenue. But the chain runs just as mechanically in the other direction: more bots, no level playing field, a sicker ecosystem, unhappy players, a business with an expiry date. The rake a bot pays is borrowed against the trust of every human at the table, and that loan always comes due.

This is the part players rarely see: when a site tolerates bots, it’s usually not a technology failure. It’s a time-horizon failure. Someone inside chose this quarter over the next decade.

Why even serious sites struggle

The sites that do commit to banning inherit a statistics problem with a cost players never price in: false positives.

Ban aggressively on behavioral signals and you’ll hit real humans. The disciplined reg with fixed sessions. The player using identical sizings because a coach drilled them. Every wrong ban is a paying customer accused of cheating, sometimes publicly. So thresholds get set conservatively, and players read that conservatism as indifference.

There’s also a real short-term price for doing the right thing. Purge a chunk of your volume overnight and tables break while the ecosystem heals. That price is exactly why only operators building for the long term pay it, and why a sudden bot purge usually isn’t new detection technology. It’s the long-term view winning an internal argument.

Build for the world that’s coming

Here’s what’s changed. Solvers made bots good. AI made them cheap. The cost of fielding a competent bot has collapsed, and it keeps collapsing.

That breaks both sides of yesterday’s playbook. Detection built on yesterday’s signals is fighting yesterday’s bots: the new ones move the mouse like a human, time their decisions like a human, and pick their tables like a smart pro. And a site designed for yesterday’s world, where botting was hard and rare, is defending the wrong perimeter entirely.

So the interesting question for the next generation of poker isn’t “how do we catch them.” It’s “what do you build for a world full of cheap AI.” Game formats that are hostile to automation. Identity that raises the cost of scale. Economics that make botting unprofitable rather than merely detectable. Detection still matters, but it has to evolve from hunting yesterday’s bots to assuming anyone can field one tomorrow.

The site of the future won’t be the site of yesterday with better filters. It will be designed, from the table up, for a world where the bots are already good.

Players see the bot that beat them. Operators see the ban queue, the false positives, and the volume graph. I’ve sat in both seats, and the lesson from both is the same: the sites that survive will be the ones that look at the volume graph and ban anyway.