Sizing configs

A full solution offers four or five bet sizes at every decision. Nobody plays that way, and the doctrine megamaster's rules are built on says you shouldn't try: one size on the flop, usually one on the turn, block plus pot on the river captures nearly all the EV. A sizing config is where you write that down — which simplified menu you practice at each decision — and the library tells you what it costs.

General defaults and Spot rules

Rules › Sizing configs. General sets the starting menu per scope: pot type (SRP or 3-bet), position (IP or OOP), role (aggressor or caller), and the decision slot — flop, turn after a bet, turn after a check-through, river after a bet, river after a check-through. Broad rows ("Everything") apply wherever a more specific row doesn't. "After a bet" means the previous street contained a bet or raise — a fact about the line, not a judgment; the flop has no previous street.

Spots are the exceptions: an ordered list of rules keyed to your texture buckets, checked top to bottom, first match wins. "Monotone flops keep all sizes; paired flops use B70; everything else uses the General default." A rule whose menu equals the default it shadows is tagged keeps default — it stops later rules without changing the answer, which is sometimes exactly what you want and sometimes a mistake, so the editor points it out. The test panel at the bottom tells you which rule fires for any flop and scope.

Built-in configs are read-only; Customize makes an editable copy. "Doctrine starter" is the doctrine as a starting point (B30 flop, B70 after a bet on the turn, B70 + overbet after a bet on the river) — which single size is right is a per-board question, and that is what the next section answers.

EV loss by sizing

Solve › Library, under a spot's rules: the priced cost of each simplified menu against the full menu, as a percentage of the pot, over your saved texture buckets. Each number is a real re-solve — the library rebuilds the tree with fewer flop sizes and solves it to the same accuracy, both players re-equilibrated — not a solution with buttons hidden. Bands follow the doctrine: under 0.5% of pot is fine for almost everyone, 0.5–1% is borderline, above 1% is unacceptable. A "+" marks an option whose solve was killed early once its cost was provably above the kill threshold: the number is a lower bound. Drag across cells to average them; "Individual boards" shows one board per row.

A config's page also prices the config itself on any solved spot: its resolved flop menu per board joined to that board's priced options — mean cost, worst board, and how many boards are unpriced. Unpriced is counted, never assumed free.

Using a config in training

Pick a sizing menu when you create a drill or start free play. Your bet buttons are restricted to the menu (check, call and fold always stay; a menu with no matching size in the tree keeps the nearest one rather than silently turning the spot check-only), and every decision shows the menu in force and — when the library has priced this exact simplification on this board — what it costs.

How it is graded, honestly. The full solution is the reference, renormalized over the sizes you were allowed: "buttons hidden". That is not the same as a strategy solved for the restricted menu, and the trainer labels it so on every decision rather than pretending otherwise. Solving restricted trees per config is library work; the priced cost shown beside each decision is what that would change.

Nothing is ever applied by default. A drill or session names a config; no other surface reads one.