FIZZ SCORE

WHAT IT IS

Fizz Score is an advisory opinion on the decisions you made, computed after the game from game-theory-based models.

Think of it as a second opinion, not a ruling. The models behind it are approximate by design, so every grade is an estimate of how your play compares to a strong baseline strategy — never the final word on what was right.

YOUR HEADLINE NUMBER

Your rating is the Fizz score, from 0 to 100 — higher is better. It measures how close your decisions are to GTO play: how many big blinds per 100 hands they gave up compared with the model (bb/100 given up). Giving up nothing scores 100; every Fizz score on the site, from your rating to each game in your history, uses this one scale.

A game’s score is your average over that game’s graded decisions, so a short game can land anywhere — a few clean hands score 100, one costly call can sink it. Your rating weighs all your games together.

The score splits into five equal GTO-accuracy levels:

• Elite — score 80–100, under 10 bb/100 given up: close to the limit of what the model can measure.
• Strong — score 60–80, 10–20 bb/100 given up: better than every one of the site’s bots with 15+ big blinds.
• Solid — score 40–60, 20–40 bb/100 given up: where the site’s tight, ABC and TAG bots land.
• Developing — score 20–40, 40–60 bb/100 given up: worse than a typical tight, ABC or TAG bot.
• Learning — score 0–20, 60 or more bb/100 given up: gives up as much as the calling-station bot, or more.

The cut-offs come from grading the site’s own bots exactly the way your decisions are graded. Only the calling-station bot stands apart on this measure; the tighter bots all land around Solid, even though they beat each other at the table. A level is shown once at least 40% of your decisions are graded.

GTO accuracy is not the same as winning. Against weak opponents — or the site’s bots — playing to exploit their mistakes often wins more chips than GTO play, and that counts against this score. Your actual results are the win rate on the Stats tab.

Your bb/100 given up comes with a 90% range: with few hands it is wide, and it narrows as you play. It assumes the model’s estimates are right, so the model’s own error comes on top. The share of decisions that could be graded is shown next to it — decisions the model can’t evaluate are counted at your average, not as perfect play. Hands the table played for you after a timeout are never graded.

THE VERDICT BANDS

Each graded decision gets one of four labels, based on how much the model estimates the action cost you, as a share of that pot. Because the labels are relative to the pot, a blunder in a small pot can cost fewer chips than a mistake in a big one — the BB figure next to a hand or a mistake is what it really cost, and it is what your rating adds up:

• ✓ acceptable

  Estimated cost up to 10% of the pot. Differences this small are within the model’s own margin of error, so they are never flagged.

• ~ inaccuracy

  Estimated cost of 10–30% of the pot. Worth a look next time the spot comes up, not a worry.

• ✗ mistake

  Estimated cost of 30–80% of the pot — the model found a clearly better option.

• ‼ blunder

  Estimated cost above 80% of the pot. A large cost by the model’s estimate — still an estimate, and in unusual spots the model can be wrong.

KNOWN LIMITS

• The advice is approximate by design — it comes from simplified models of the game, not a complete answer.
• Pots with three or more players have no single best strategy, even in theory. Advice there is an informed estimate.
• Some unusual action sequences aren’t evaluated at all. When that happens, the decision says so — with the reason — instead of getting a guessed grade.
• The model counts chips, but in a tournament finishing places matter too. Late all-in decisions with three or four players left — where a medal or league points can change the right play — are therefore left ungraded (“tournament stage”).
• The Fizz score and bb/100 given up come from the same per-decision estimates, so they carry the same limits. Small samples swing a lot — trust the range, not a single game.

WHEN ADVICE APPEARS

Only after a hand is over.

Nothing here runs during play. At the table you get no hints, no live advice and no grading — analysis happens once the game is done.