Deblunder study notebook
What is a blunder in chess? Blunder vs mistake vs inaccuracy
A blunder is a move that gives away a large part of your chances in one decision: material, a won position or a forced mate. Engines label it by how much the evaluation drops. The label tells you where to look, not why it happened.
The short definition
In chess notation a blunder is marked with ?? after the move. It usually loses material, allows a tactic or misses a forced defense, and it changes the expected result of the game. A mistake (?) is smaller: the position gets clearly worse but the outcome often stays open. An inaccuracy (?!) is a weaker move that keeps the position playable.
These are labels for the consequence of a move, judged by an engine after the game. Two players can make the same blunder for different reasons: one missed a check, another moved a piece that was blocking a line. Training works on the reason, so the label is only the first step.
How Chess.com classifies a blunder
Chess.com's help center describes an expected points model: 1.00 means always winning, 0.00 always losing and 0.50 even. Each move is judged by how many expected points it costs. An inaccuracy costs 0.05 to 0.10, a mistake 0.10 to 0.20 and a blunder 0.20 or more.
The same article adds Miss (failing to punish an opponent's error), Great and Brilliant moves, and notes that the judgment of winning, equal or losing positions varies by rating. That is why a move can be a blunder in one review and a mistake in another tool.
How Lichess classifies a blunder
Lichess analysis is powered by Stockfish. Its open source code converts each evaluation into winning chances on a scale from -1 to +1 and compares the position before and after your move. A drop of at least 0.1 is an inaccuracy, 0.2 a mistake and 0.3 a blunder. Separate rules describe mate sequences, such as allowing an unavoidable checkmate or losing a forced one.
Because the scale is about chances, losing a pawn in an equal position can count more than losing a piece in a position that was already completely won. The classification measures practical damage to the result, not the material count.
Why the label is not the lesson
A blunder report answers one question: where did the evaluation fall? It does not tell you what you were thinking, which threat you did not consider, or whether the same oversight appears in your other games. Players who only read the labels often fix the move and repeat the cause.
A more useful habit is to write one sentence per blunder: what you intended, which reply you missed, and which relation on the board made it work. For example: 'I captured with the bishop, but it was the piece blocking the e-file, so my rook on e8 was exposed.' That sentence is something you can train.
From blunders to a training plan
Collect the positions just before your blunders across several games of the same time control. Group them by cause: unseen checks, pieces that were defending or blocking, back rank weaknesses, moves played too fast. The cause that repeats is your priority, even if a single dramatic blunder feels more important.
Deblunder automates this step: it analyzes completed games from Chess.com, Lichess or PGN, explains the cause of each error with pieces and lines when the board proves it, marks the rest as unknown, and turns recurring causes into training with your own positions.
Frequently asked questions
What does ?? mean in chess?
The ?? annotation marks a blunder: a move that seriously worsens the position, usually by losing material, allowing a tactic or missing a forced defense. A single ? marks a mistake and ?! a dubious move or inaccuracy.
Is a blunder the same on Chess.com and Lichess?
Not exactly. Chess.com uses expected points (a blunder costs 0.20 or more), while Lichess uses winning chances on a -1 to +1 scale (a drop of 0.3 or more). Engine depth and settings can also change the result, so the same move may receive different labels.
How many blunders per game is normal?
There is no universal number. It depends on rating, time control and how sharp the games are. A more useful measure is whether the same cause keeps appearing in your next games of the same time control.