Notes from the felt · June 2026

Club Omaha: where the bots actually sit.

Here's the thing about club Omaha — the ecosystem is different. The agent layer matters more than the algorithm.

If you're picturing PLO bots grinding away in some big-name online room, recalibrate. That's not where the game lives in 2026. The real action — and the real bot activity — is in private mobile-app clubs: PPPoker, UPoker, PokerBros, and a long tail of regional skins running on the same underlying tech. The clubs are invite-only. Buy-ins are settled in cash or crypto between trusted parties. The lobby is a Telegram group. The dealer is an app on someone's phone.

So what's an "agent" and why does it matter?

In club poker, you don't deposit on the app. You deposit with an agent. The agent is a person — usually with a personal network of players — who runs accounts inside one or more clubs, manages chip transfers between players and the club, and earns a cut. From the player side, you wire money to your agent, they top up your in-app chips, and at the end of a session your agent settles up.

That layer is what makes the whole ecosystem run, and it's what makes bots viable. The agent vouches for the seat. If your bot is registered through a trusted agent in a softer club, the club operator has a financial interest in not looking too hard. Bots that get caught aren't usually caught by detection software — they get caught by other agents complaining their players are losing too consistently to the same handful of seats.

Why PLO specifically?

Two reasons. One, PLO games have huge variance, which means the cluster of recreational players — many of whom genuinely believe they're "running bad" for months — are slow to figure out they're being beaten. Two, the higher-stakes PLO games at private clubs have real liquidity. You can find 25/50 and 50/100 PLO running across a half-dozen clubs at any given evening hour in the US-friendly timezone. That liquidity makes a bot operation economic in a way that small-stakes hold'em isn't.

The flip side: clubs run shorter games, the player pool is smaller, and your bot's identity in the lineup is fragile. One observant reg with a hand-history converter and a few weeks of data can map a bot's tendencies. That's why operators rotate seats, rotate clubs, and run multiple personas — not because the algorithm is weak, but because the surveillance environment at the club level is human, not algorithmic.

How do you actually get into a club?

You don't sign up. Someone vouches. An existing agent or club owner adds you through their referral chain, you pass a brief know-your-account-bankroll check (mostly: can you actually pay if you lose), and you get an in-app ID and a Telegram contact. From there, the game-selection skill matters as much as the technical skill. The same bot in two different clubs can be a money-printer or a slow leak, depending on the lineup.

What changes for the bot designer?

Three things, mostly:

• Detection model is human. You're not dodging a casino's anti-collusion AI. You're dodging a sharp reg with a lot of time on their hands. Different problem.
• Action timing matters more. Mobile clubs are played on phones. Players take time. A bot that snaps every decision in 0.4 seconds stands out worse than one that occasionally "thinks" for 8.
• Persona stability matters. Your bot needs to play roughly the same way across sessions, but not identically. Drift the strategy slightly. Be a player, not a function.

Back to the homepage, or the math side of how these bots think.