Everything below ships in the app today.

01 · Games

Multiway first, Omaha first

• Pot-Limit Omaha, first-class. Four-card PLO for 2–6 players, preflop through river. Multiway is the design center, not an afterthought bolted onto a heads-up engine.
• PLO5. Five-card Omaha for turn and river spots.
• Omaha hi-lo (8-or-better). Full hi-lo split evaluation.
• Multiway NLHE at exact 1326-hand granularity. Hold'em with standard range syntax (QQ+, AKs, ATo+) — every hand tracked individually, no hold'em abstraction.

02 · Tournament

Chips are not equity

• ICM, exact-validated. Malmuth-Harville tournament equity validated against brute-force enumeration to < 0.2% of the prize pool — see the methodology. Bubble factors are reported automatically for every ICM spot.
• Large-field MTT. Model the rest of the field's stacks so equities reflect the real tournament, not just the table.
• PKO bounties, with one-click A/B. Bounty fraction and carry-over, including mystery bounties as expected draw value. “Compare without bounties” re-solves the same spot with bounties zeroed and shows the per-seat EV delta — how much of the strategy is bounty pressure.
• Future game simulation (FGS). Expected-blinds future-game adjustment for decisions whose real value depends on the hands after this one.
• Rake models. Percentage with cap, applied to the called pot only, plus a no-flop-no-drop mode for rooms that don't rake preflop fold-outs.

03 · Analysis

Interrogate the solve

• Strategy browser. Click through the tree while training runs: a line strip of colored action and card chips, ←/→ runout stepping, and a sortable all-runouts table at every chance node — action mix and per-card EV for every next card, grouped by card class (pairs the board / flush card / straight card / brick), and filterable by hand pattern (AAxx, KKxx:ds, a specific card).
• Per-hand EVs with error bars. Monte-Carlo EV per action for any hand at any node, reported as EV ± standard error — you always know how much to trust a number. Toggle an EV column for the top hands in range.
• Node locking with auto-rebalance and baselines. Per-action sliders seeded from the current mix; moving one rescales the rest. Lock the node, resume solving, and the other seats learn the maximally exploitative counter. A baseline snapshot is taken at the first lock, so after the re-solve you read the answer directly: per-seat EV before, after, and the delta. Remove all locks in one click. Locks save into the spot file and work from the CLI too.
• Incentives. Soft EV bonuses at chosen nodes — tilt a player toward a line without hard-locking frequencies.
• Sampled exploitability. End-of-solve best-response exploitability estimate, so convergence is measured, not assumed.
• Strategy simplification. Round displayed mixes to a 25% grid, and measure the EV cost of rounding the strategy to coarser frequency steps.
• Equity calculator. Board + ranges equity from the GUI or plo_equity on the command line.
• Practice mode with lasting stats. Play the spot against the solution at a real table view. Every decision gets a frequency grade and an EV verdict for your dealt hand — how many bb the chosen action gave up against best. Drill from any node (“Practice from here” deals hands from the ranges that actually reach it), and results persist across sessions: blunder rate, average EV loss, your worst street, and a per-spot breakdown.

04 · Workflow

Built for study routines, not demos

• Start from a template, a pasted hand, or another solve. Eight built-in templates (SRP, 3-bet pot, bomb pot, final-table ICM, PKO, river study…) open a ready spot instead of a blank form. Paste a hand history and the spot form prefills itself for the last street you played. Or spawn a child solve from any node of a finished solve — “Solve from here” carries the reach ranges, pot, stacks, and rake over automatically, so a preflop solve feeds its own postflop studies with no copy-paste.
• Spot files. Every solve is a small, versionable JSON file — players, board, stacks, sizes, ranges. Reproducible by construction.
• Checkpoints and solution packs. Solves checkpoint and resume. A spot file plus its checkpoint is a shareable solution pack: the app opens it in a read-only viewer mode, no retraining possible by accident.
• Batch solving + aggregation reports. Give one spot a board list (or generate a texture subset — paired, monotone, connected — from the 1,755 canonical flops, correctly weighted) and solve them as a queue: per-board checkpoints, converged boards skipped on resume. Then read the aggregate: strategy per board, texture-group means, and the outlier boards that deviate most from their group. In the apps and the CLI (plo_cli, plo_boards, plo_agg) with interchangeable artifacts.
• Node deep links and one-click opening. Copy a prism:// link to the exact node you're looking at; the recipient's app walks straight to it. Spot files and solution packs open by double-click or drag-and-drop.
• Tiered memory presets. The laptop preset solves postflop spots in minutes on ordinary hardware; the box preset scales bucket counts to workstation RAM. A live cost meter shows tree size and implied memory before you build.
• Custom trees. Size menus as pot fractions, big-blind targets (2.5bb), raise multiples (3x), or geometric sizing; rules selectable per street, seat, and facing/donk context. Raise caps, min-bet control, per-seat stacks with side pots, all-in thresholds, antes, and single/double/button straddles.

Feature descriptions match the shipped user guide. The full manual is bundled with the app (Help → User Guide).