Inaccuracy vs mistake vs blunder: what's the difference?
An inaccuracy, a mistake and a blunder are the same thing at three sizes: a move that throws away part of your winning chances compared with the engine's best move. On Chess.com an inaccuracy costs 0.05–0.10 expected points, a mistake 0.10–0.20 and a blunder 0.20 or more; Lichess uses drops of 5, 10 and 15 percentage points of winning chance.
The label says how much a move cost you, not why. That's the part people get stuck on, so below are the official cutoffs from both sites and one real example of each, taken from my own Chess.com games (I'm nacsol3, an adult improver who went from 631 to 1385 in rapid).
Inaccuracy vs mistake vs blunder on Chess.com and Lichess
| Label | Chess.com (expected points lost) | Lichess (winning chance lost) |
|---|---|---|
| Inaccuracy | 0.05 to 0.10 | 5 to 10 percentage points |
| Mistake | 0.10 to 0.20 | 10 to 15 percentage points |
| Blunder | 0.20 to 1.00 | 15 percentage points or more |
| Losing a forced mate | Not a separate rule | Its own rule: inaccuracy if you're still above +9.99 afterwards, mistake above +7.00, blunder otherwise |
Sources, both checked October 26, 2026:
- Chess.com: How are moves classified? (Help Center, dated February 9, 2026). Expected points run from 0.00 (certain loss) to 1.00 (certain win), with 0.50 as even. The model estimates them from the player's rating and the engine evaluation. The same table puts Best at 0.00, Excellent at 0–0.02 and Good at 0.02–0.05.
- Lichess: the accuracy page gives the formula that turns an evaluation into a win percentage. The thresholds live in the open-source Advice.scala as 0.1, 0.2 and 0.3 on a −1 to +1 scale, which works out to 5, 10 and 15 points on a 0–100% scale. The forced-mate row comes from the same file.
Two things follow from the table. First, the scales aren't identical: Chess.com's Blunder line starts higher (0.20) than Lichess's (15 points), and Chess.com also weighs your rating. So the same move can get a different label on each site. Second, a big number swing isn't always a big label. Going from +12 to +9 barely changes your chances of winning: by the Lichess formula both are above 96%, so Lichess doesn't label that move at all. The guide to reading Stockfish analysis explains how to turn the evaluation numbers into winning chances yourself.
A real inaccuracy: taking the wrong thing
March 22, 2025. I'm White (1307) against Ezequiellk (1337) in a 10-minute game (game on Chess.com). I traded bishops on d6. Stockfish preferred grabbing the pawn on b7 with the queen, which also hit the undefended rook on a8.
My game analysis has the evaluation going from +1.11 to +0.11, which is 60.1% to 51.0% winning chance: a drop of about 9 points. That's an inaccuracy on the Lichess scale. Nothing was lost on the board. I just let a small edge slip, which is exactly what an inaccuracy usually is.
A real mistake: a quiet move that drops a pawn
August 6, 2024. I'm White (993) against fosenheimen (1016) (game on Chess.com). Move 32, rooks and knights on the board. I put a rook on e1. Stockfish wanted my king on b2, attacking Black's rook on a3, which had no defender.
After my rook move, the engine's best reply for Black is to take the pawn on c3 with the knight. The evaluation went from −0.37 to −1.68 (46.6% to 35.0%), a drop of about 12 points. That's the typical mistake: the position goes from equal to clearly worse, but not lost.
A real blunder: castling into a fork
December 31, 2025. I'm White (1325) against CSW72 (1283) (game on Chess.com). I castled queenside. That put my rooks on d1 and h1, and Black's knight on e4 could jump to f2 and attack both at once (a fork is one piece attacking two things at the same time).
My opponent found the knight jump right away. The evaluation went from −0.32 to −6.62, from 47.1% to 8.0% winning chance: a drop of 39 points, more than twice the Lichess blunder line. I still won that game, because my opponent went wrong later. A blunder is a verdict on one move, not on the result.
How to use the three labels
- Blunders first. They cost the most. Look for the pattern: a piece left undefended, a fork, a missed check.
- Then mistakes. They're often a plan problem, like the rook move above that ignored a loose enemy piece.
- Inaccuracies last. At beginner level, fixing them gains little until the blunders are under control.
If Chess.com shows you the label but not the reason, here's why Game Review isn't explaining your moves.
How Chess Coach labels moves
Chess Coach, a connector for Claude and ChatGPT, runs Stockfish on our own server and uses the Lichess win-percentage formula with the same 5, 10 and 15 point thresholds. It doesn't label a move that was the engine's top choice. The examples on this page come from those analyses.
The difference is the explanation. After a review, your assistant shows the key moments on an interactive board, with a red arrow for your move and a green arrow for the better one, and explains in plain words what the move missed. Stockfish finds the problem and the assistant explains it, which is the split described in ChatGPT vs Stockfish. The free plan includes one in-depth review a day. Connect Chess Coach and ask it to review your last finished game.
Frequently asked questions
Is an inaccuracy bad in chess?
Not very. An inaccuracy costs a small slice of your winning chances: 0.05 to 0.10 expected points on Chess.com, or at least 5 percentage points on Lichess. On their own they rarely turn a game around.
Which is worse, a mistake or a blunder?
A blunder. On both Chess.com and Lichess the order is inaccuracy, then mistake, then blunder, from smallest to biggest loss of winning chances.
Why is the same move a blunder on Chess.com but a mistake on Lichess?
The two sites use different models. Chess.com's Expected Points model includes the players' ratings, and its Blunder line starts at 0.20; Lichess uses only the engine evaluation, and its Blunder line starts at 15 percentage points of winning chance.
What is a Miss on Chess.com?
Chess.com's Help Center describes a Miss as failing to take advantage of your opponent's mistake. It isn't defined by an expected points range like Inaccuracy, Mistake and Blunder.
Does a blunder always lose the game?
No. A blunder means the engine thinks your chances dropped sharply, not that the game is over. In the blunder example on this page I lost a rook to a knight fork and still won, because my opponent went wrong later.