What Makes a Chess Move a Blunder?
What makes a chess move a blunder?
A chess move is a blunder when it lowers the mover's win chance by 20 percentage points or more in a single move. BlunderLoop analysis policy v1 grades every move by its win-chance drop: a drop under 2 points is best, 2 to under 8 points is an inaccuracy, 8 to under 20 points is a mistake, and 20 points or more is a blunder.
Win chance, not centipawns
Engines report evaluations in centipawns (hundredths of a pawn), but a fixed centipawn loss does not hurt equally everywhere. BlunderLoop first converts every evaluation to a win chance on a 0-100 scale using the lichess-standard sigmoid: Win% = 50 + 50 x (2 / (1 + e^(-0.00368208 x cp)) - 1). Centipawn scores are clamped to +/-1000 before conversion, and forced-mate scores map to the near-extremes of the scale.
Because the curve is a sigmoid, the same centipawn swing is punished heavily around equality and barely at the extremes. That matches how games are actually won and lost: near equality every fraction of a pawn changes the likely result, while in a decided position extra material hardly moves the outcome.
The four grades of policy v1
A move's grade is the difference between the mover's win chance before and after the move, measured in percentage points. The thresholds are fixed by analysis policy v1 and recorded with every report.
| Grade | Win-chance drop |
|---|---|
| Best | under 2 points |
| Inaccuracy | 2 to under 8 points |
| Mistake | 8 to under 20 points |
| Blunder | 20 points or more |
Why hanging a piece in a lost position is not a blunder
Dropping three pawns of material in an already-lost position (about -600 centipawns) only moves your win chance from roughly 10% to roughly 3.5% — a small drop, so policy v1 does not call it a blunder. The game was already decided; the move changed little.
The same 300-centipawn swing through equality — say +150 to -150, roughly 63% to 37% win chance — turns a playable game into a lost one. That 26-point drop crosses the 20-point bar and is graded a blunder. Grading by win-chance drop keeps the label attached to moves that actually changed the result, which is what makes the drills built from them worth training.