Guides

PLO: pot-limit Omaha blueprint and postflop solver

The desktop Engine has a PLO mode: a 6-max 100 bb preflop blueprint for pot-limit Omaha and a heads-up postflop solver for it. Open it from the mode menu in the toolbar (PLO). The Strategy tab has the same layout as in Hold'em: the grid with its rail, the seats strip, the action buttons with their EV, the hero's hand and the Strategy detail tabs. Solving and training run on your PC. You can also browse the published PLO blueprint in Cloud, using your existing Browse allowance.

PLO requires PulseGTO 1.0.71 or later. The seven-day free trial includes it, with Hold'em, and needs no payment card: sign in and get an activation code from your dashboard. After the trial PLO needs a PLO Engine or Both engines licence; see pricing for the plans.

1. Get the blueprint

• Download PLO 6-max 100 bb — GG Rush & Cash from the Download page, a .plobpz file of about 1.95 GB. In PLO mode, open Preflop Settings: the file is listed from your Downloads folder, and Install unpacks it (about 8.5 GB) and puts it in use.
• The game: 6-max, 100 bb, blinds 0.5/1, rake 5 % capped at 1 bb, no flop no drop. Opens and re-raises are pot-sized, up to the 5-bet; the small blind may complete; at most three players see a flop.
• It was trained for 10 hours (1.5 billion deals) at the small abstraction level. Opens: UTG 17.4 %, HJ 22.1 %, CO 30.3 %, BTN 47.2 %, SB raise 29.6 % and complete 19.7 %.

2. Preflop

• The blueprint has one strategy for each of the 16,432 PLO hand classes (hands that differ only by a relabelling of suits play the same); all 270,725 hands map onto them.
• The 13×13 grid groups the hands. With TOP2 (the default) a hand counts in the cell of its two highest cards: AKxx in AK (suited when a king shares the ace's suit), QQxx in QQ, 9876 in 98. With ANY2 it counts in the cell of every pair of its cards, so AA72 also shows in 72o: a cell then tells how hands holding those two ranks play. The switch is on the rail, under COMBO / REACH / FULL.
• Walk a line with the action buttons; the seats strip shows each decision and a click on a played one goes back to it.
• Hovering a cell opens the cell window: its hand classes with their mix and EV. A click pins it; click the cell again or press Esc to close it. A pinned window lists every hand of the cell (scroll for more) and finds hands by their cards: type AcAdKcJs for one hand, Ac Ad for every hand holding both, or AAK for ranks of any suit. The Strategy detail tabs list every hand class with its mix and EV (Table View), by rank, along the range (Graph View) and as an EV curve.
• The cells button on the rail cycles the grid's views. TOP2: a hand's two highest cards, suited when those two share a suit. DS, S+D and S+D+T: the same cells with GTO Wizard's suited side, so a hand sits above the diagonal when it is double-suited (DS), single- or double-suited (S+D, Wizard's default) or also triple-suited (S+D+T). ANY2: every pair of a hand's cards.
• Type a hand under the action buttons (the Hero field): every cell holding two of its cards pulses, and the Table View pins the hand on top.
• Purify: as in the Hold'em blueprint view, a hand class's actions under 2 % are dropped and the rest renormalised, in the grid and in the ranges passed to the flop. The setting is shared with Hold'em (Settings → Engines → Blueprint purify) and also sits in PLO's Preflop Settings; 0 shows the raw averages.
• EVs are the blueprint's training averages per hand class, in bb.

3. After the flop

• At a heads-up flop line, pick the three flop cards (or Random): the blueprint's own postflop play on that flop opens at once, approximate, with hands in buckets and one bet size (66 % of the pot, one pot raise). Random on the turn and river columns deals those cards. EQ shows both ranges' equity.
• The blueprint's postflop view shows EVs for both players: the EV grid, the EV on each action button and both EV curves. They are estimates for each hand bucket, averaged over all the flops of the line; Re-solve estimates EVs for the selected street without card abstraction, subject to the solve's reported accuracy.
• The Table View lists each hand's EV, EQ (its all-in equity against the other player's range) and EQR (EV ÷ (EQ × pot)). On the turn and river EQ and EQR fill in by themselves; on the flop, switch the rail to EQ to compute them (a flop takes a few seconds).
• Run solves the flop with the PLO solver from both players' blueprint ranges and the sizes in Postflop Settings. A flop solve keeps turn and river decisions in hand buckets, so it is approximate; Re-solve on a turn or river node solves that street without card abstraction, with every hand separately and every river card. The reported solve accuracy still applies.
• Re-solve keeps the tree the node is in, as in Hold'em: in the blueprint's postflop view the blueprint's own sizes (one 66 % bet, pot-size raises) and its rake, in a solve that solve's sizes. Change node re-solves the same street with the sizes in Postflop Settings instead.
• Re-solve and Change node work on every turn and river decision, also inside a re-solve. Each re-solve starts at the street's card with the ranges the solve on screen reaches it with. For one consistent line, go forward: re-solve the flop from the blueprint view, then the turn inside that solve. A turn re-solve already plays the river exactly.
• Postflop is heads-up: three-way flops show the ranges only.

4. Train your own blueprint

Preflop Settings → Train a new blueprint sets the stack, rake, open and re-raise sizes, the number of raises, players to a flop, the small blind's complete, the postflop bet size and the Abstraction level: small, medium or high, as in Hold'em. Training uses all cores at below-normal priority, can be stopped and resumed (a resumed blueprint keeps its level), and the Strategy tab shows it as it learns.

• Postflop EVs (collect or off) sets whether training also estimates the postflop EVs the blueprint view shows. They need the extra RAM in the table below; off saves it.
• In the list of your blueprints, Continue trains one further with its own settings, and Compute EVs estimates its postflop EVs as it is, without changing its strategies. Both use the training time and threads set in the form.

• After the flop, a hand is described by its made strength against every holding the opponent can have (kickers, nut flushes against lower flushes, nut straights and the board's threats all count), by what the next card does to it (outs to the nuts and to strong hands; for a made hand, the cards it survives), by backdoor flush draws on the flop, and by the share of the opponent's strongest holdings its cards block.
• On the flop and turn, hands are clustered on strength and next-card outs, as the Hold'em blueprint clusters hands, and each cluster is split by blockers. On the river, hands are binned by strength and blockers.

The last column is measured on solves of 3 flops, 6 turns and 6 rivers with the blueprint's own ranges: the EV a player gives up at a node when all the hands of a bucket play one strategy. It covers one node on one board; a blueprint also shares a strategy row across all the boards of a line, which an exact re-solve removes. Higher levels need more training for the same number of samples per bucket: medium about 3 times the deals of small, high about 7 times.

Scope

• The download is 6-max 100 bb; for other stacks, rakes or sizes, train your own blueprint (section 4).
• Up to three players see a flop; a fourth player may fold or raise.
• Postflop solves are heads-up. Hand-history import is a Hold'em feature for now. The published PLO blueprint can be browsed in Cloud; custom PLO solving and training run in the Windows desktop Engine.