Chess glossary

Every game review labels your moves: best, strong, good, inaccuracy, mistake, blunder. Here’s exactly what each one means and how it’s calculated.

Centipawn

A hundredth of a pawn, the unit a chess engine uses to score a position. +100 centipawns (“+1.00”) means the position is worth about one extra pawn for whoever's ahead. Engines report scores this granular because a move can be better or worse by far less than a full pawn.

Centipawn loss

How much worse a position got, in centipawns, because of the move actually played, compared to the engine's best available move at that point. If the engine had the position at +50 and your move drops it to −50, you lost 100 centipawns of evaluation, roughly a pawn's worth of advantage, even though no pawn was captured. This is the single number almost every other term on this page is built from.

Win percentage / win chance

Centipawn loss alone can be misleading: losing 100 centipawns in an already-decided position matters far less than losing 100 centipawns in a close game. Win% converts an engine score into an estimated probability of winning from that position (via a standard sigmoid curve), so a move is judged by how much it changed your actual chances of winning, not just the raw evaluation.

Best★

The move the engine considers strongest, or any move that delivers checkmate. No win% was given up.

Strong✦

Not the engine's top choice, but close enough that the difference is negligible: a win% drop of 2% or less, measured reliably. Often the position simply had more than one good option.

Good

A solid, reasonable move. Win% dropped by 10% or less. This is the default label; most moves in most games fall here.

Inaccuracy?!

A move that gave up more than it needed to: a win% drop between 10% and 20%. Not a losing mistake on its own, but a step in the wrong direction.

Mistake?

A clearly worse move, costing 20 to 30% win%, or a bigger drop that gets capped down to a mistake because the position was still comfortably winning afterward (see “Severity cap” below).

Blunder??

A move that cost more than 30% win% AND left the position at less than roughly +3 pawns for the mover. This second condition matters: a big evaluation swing inside an already-crushing position isn't really a blunder, because the result was never actually in danger.

Severity cap

A large drop in engine evaluation only earns the “blunder” label if it actually put the result at risk. If your position was still winning by three pawns or more after the move, the label is capped at “mistake”. The idea is that “blunder” should mean something concrete: you jeopardised a game you were winning, not that a number moved on a graph.

Accuracy

A game-level (or player-level) summary of how closely someone's moves matched engine-recommended play across a whole game, usually expressed as a percentage. It's built up from the same win% math as individual move classifications: fewer and smaller drops across a game means a higher accuracy score.

Engine

The chess-analysis program that calculates the objectively best moves and evaluations for a position by searching millions of possible continuations (Socratic Chess uses Stockfish). Every claim a Socratic Chess coach makes about a position is checked against a live engine search, not guessed.

Evaluation (eval)

The engine's numeric judgment of a position, in centipawns (or as a forced mate in N moves). Positive means White is better, negative means Black is better. A move's centipawn loss is just the change in evaluation it causes.

Principal variation (PV)

The engine's predicted best sequence of moves from a given position, not just the next move, but the likely line that follows if both sides continue playing well.

Want to see these labels applied to a real game? Bring over your own chess.com games, or open a curated classic to see a full annotated review.