Notes from the felt · June 2026

You want a PLO bot. Let's talk about what that actually means.

A no-limit bot ported to Omaha is a no-limit bot losing money in Omaha. Four hole cards aren't a tweak. They're a different game.

Look — if you've ever sat in a 5/10 PLO game at a private club, you already know the bots are there. The only question is which seats. What people get wrong is assuming a bot built for hold'em can just be retrained on PLO hand histories and shipped. It can't. The shape of the decision tree is different, the equity distributions are different, and the way value moves around the pot is different. If your bot doesn't internalize that, it bleeds.

Why is Omaha not just "hold'em with more cards"?

In no-limit hold'em, the average all-in equity gap between the best hand and a random hand is enormous. Aces are 80%+ versus a single random hand preflop. In PLO, the strongest possible starting hand — double-suited rundown aces — is roughly 70% against a random hand, and against any reasonable opponent range it's usually a 55–60% favorite. Equities run close. Nuts change on every street.

That single fact ruins half the heuristics a NLH bot relies on. "Hand strength" stops being a stable feature. A pair of aces on a dry board is a bluff catcher, not a value hand. You have to model your opponent's hand as a distribution over 270,725 possible four-card combos, not 1,326. The state space explodes. So does the cost of running solver-style approximations in real time.

What does a PLO bot actually need to model?

Three things a hold'em bot can hand-wave that an Omaha bot can't:

• Redraw equity. You don't just have the current best hand. You have the hand that can become a better hand by the river. A wrap with a flush draw is a different beast than a wrap dry.
• Blocker structure. With four cards, your blockers are dense and asymmetric. Holding the bare ace of spades on a three-spade board means something different than holding the ace with another spade in your hand.
• Pot-limit geometry. Bet sizing is constrained. Pot, half-pot, third-pot — these aren't arbitrary buckets, they're the only legal moves at certain stack-to-pot ratios. Your bot's action space is structurally narrower than NLH, which sounds easier but actually makes deception harder.

Skip any of those and your bot looks fine on showdown stats and bleeds chips on the streets you never see.

So does a PLO bot beat good regs?

Honest answer: it depends on what you mean by "good." Against a player who treats PLO like NLH with two extra cards — flat too wide preflop, overvalues top set on wet boards, doesn't think about redraws — the bot crushes. Against a real PLO reg who understands equity realization and bet-sizing in 4-card pot-limit, the edge collapses fast. A well-built bot can be break-even-to-slightly-positive against a tough lineup, which sounds modest until you remember most lineups aren't tough.

The money in club Omaha doesn't come from out-soloing pros. It comes from sitting in the right games and not getting found out. Game selection > cleverness.

Where does this all happen?

Not on the public sites you've heard of. PLO bot activity in 2026 lives where the games live: private mobile-app clubs running on PPPoker, UPoker, PokerBros, and a long tail of regional skins. More on that here. If you want to dig into the math side, the four-card equity piece walks through the parts hold'em players usually skip.