Analyse

The Analyse tab grades the hands you actually played. For each hand it finds the solved board for that spot, walks your real line through it, and reads what the solver does with your exact two cards at every decision you made after the flop — not what it does with your range on average. Each decision gets a tier and, where the board can price it, the EV it cost you in big blinds.

Grading your hands

Press Score my hands. Scoring runs in the background with a progress bar: first it works out which hands can be graded, then it grades them. It needs no solving of its own; it reads boards that are already solved. After that the button reads Score new hands, which grades only hands that have not been looked at yet, so run it after each import.

Re-score everything checks every graded hand and regrades the ones whose solved board the library has since replaced, or that an app update grades differently. Grades that still stand are kept.

Grade preflop grades your preflop decisions against the preflop solve. It is quick and needs no solved boards. A preflop grade is a frequency grade only, so it has a tier but no bb figure.

The filter bar at the top of the app applies here, as it does on Lines, Population and Results.

Solved boards download as needed

The library stores each spot as up to 1,755 strategically distinct flops. When a hand needs a board that is in the library but not on your machine, scoring downloads it as it gets there. Your hands are not uploaded; only solved boards come down. If a download fails (offline, for example), the hand is listed as Tree not downloaded and the next scoring run tries again.

Downloaded trees are kept in the library cache so that re-grading is quick. Before a scoring run the app shows how much is cached, the cache's cap and the free space, and lets you change the cap; the cache never grows past it, and the least recently used trees are removed first. The same cap is under Solve › Library › Storage.

What it grades

• Heads-up flops only. Every solve in the library is heads-up, so a flop three or more players saw is refused, not graded against the nearest thing.
• The preflop has to be the tree's own. Every tree starts the same way: one player opens, the other calls or raises, and nobody else put a chip in. A squeeze, a cold-call or a limp by a player who then folded, a limp before the raise, a straddle or missed blind, or a later seat raising a player who had already posted or limped all leave a pot no tree starts from, so those hands are refused. Hands from a heads-up table are graded on the heads-up trees: the single-raised, 3-bet and 4-bet pots the button opens. Any other heads-up pot is refused, and a six-handed tree is never used for a heads-up hand.
• Sizes close enough, marked when not exact. The library has one open size and one stack depth per spot. A hand whose open (or last preflop raise) is within 20% of the tree's, and whose starting stacks are inside the tree's range, is graded against it: 75–150bb for a single-raised pot (Settings › Single-raised stacks changes it, for the Lines tab too), within 25% of 100bb for a 3-bet or 4-bet pot. If it is not an exact match the EV figure carries a ≈ mark and the hand says nearest tree, not an exact match: read it as approximate. Further off than that, the hand is refused.
• Your flop, turn and river decisions. River decisions are graded after the dealt river is re-solved, because library boards are stored without their rivers.

Reading the results

Your score

Four figures: the share of graded decisions where you took an action the solver plays (Best or Correct), how many decisions were graded, the bb lost per 100 graded hands, and the bb lost per mistake. Below them, every graded decision on the trainer's ladder:

• Best — the action the solver takes most often with this hand.
• Correct — an action the solver genuinely mixes in (3.5% of the time or more).
• Inaccuracy — the solver takes it, but rarely.
• Wrong — the solver never takes it here.
• Blunder — never taken, and losing at least 3% of the pot.
• Ungraded — no strategy could be read at that node. Shown, and left out of the percentages rather than counted as right or wrong.

These are the same tiers the trainer uses; see Trainer score.

What it found

Per street: decisions graded, mistakes, EV lost in bb, and EV lost per decision. A mistake is a decision that cost more than 0.1bb against the best action for your hand; smaller differences are solve noise. All figures are in big blinds at the trees' own scale.

Costliest hands

Your graded hands, most expensive first, filterable by street. Each row shows the date, spot, your cards, the board, what you won in the hand, the EV lost and the number of mistakes. Open a row to see every decision: its tier, what you did and how often the solver does it with your hand, the best action, the kind of mistake (folded too much, called too wide, missed a bet, missed a raise, wrong size, too aggressive) and the EV lost. Open node opens that exact node in the solver viewer; Replay this hand opens it in the replayer, which shows the same grades beside your actions.

Coverage

At the bottom: how many hands you saw a flop in, how many of those were graded, how many preflop decisions were graded, and how many solved boards and spot templates are on this machine. Below that, every reason a hand could not be graded, with counts, and for some reasons the spots behind them. Hands you folded before the flop are not counted: there was nothing of yours to grade after the flop.

Decisions that are shown but not counted

• Off tree. Once you take an action your hand never takes in the solve — or takes less than 0.5% of the time, which the app shows as "solver 0%" — the solve has no meaningful strategy for your hand below it. That action is still graded; it is the mistake. The decisions after it are labelled off tree and left out of every total. The Lines map docs explain the mechanism.
• No price. Boards solved before 21 September 2026 can price river decisions but not flop or turn ones. Those decisions keep their tier and read no price instead of a bb figure. See Why some EV comparisons are missing.

Limits

• It can only grade spots the library has solved. Some templates have no boards yet, and multiway pots, limped pots and unusual preflop lines have no template at all. The coverage report says how much of your database that leaves out, and why.
• A grade compares one decision with the solver's equilibrium. It does not account for how your opponents actually play; the Lines map sets your lines against the pool's response.
• A figure graded against the nearest tree rather than an exact one is approximate, and is marked.