You've probably seen it on a post-game report: an accuracy percentage next to a number like "avg. centipawn loss: 34." Most players glance at it, shrug, and go straight to the move list. That's a mistake — centipawn loss is one of the most useful numbers in chess analysis, because it tells you the truth about how you played, independent of whether you actually won the game. You can win on time scramble blunders just as easily as you can lose a near-perfect game to one bad moment — the scoreboard doesn't separate the two, but centipawn loss does.
What is a centipawn?
A centipawn is simply 1/100th of a pawn. It's the unit chess engines like Stockfish use to measure who is winning and by how much, because pawns alone aren't precise enough. An evaluation of "+100" centipawns means the engine thinks the position is worth about one extra pawn for White; "-250" means Black is up roughly two and a half pawns' worth of advantage. It's just a fine-grained ruler for position quality, nothing more mysterious than that.
How centipawn loss is calculated
Centipawn loss measures the gap between the engine's best move in a position and the move you actually played. The engine evaluates the position before your move, evaluates it again after your move, and the difference — how much worse the position became compared to playing the top engine choice — is your centipawn loss for that move. Play the best move and your loss is 0. Miss a free piece or walk into a tactic, and that single move might cost you several hundred centipawns. Average this number across every move in the game and you get your game's average centipawn loss — the figure most analysis tools, including ChessOmni, show you at the end of a review.
What counts as a good or bad score
As a rough guide for the 800–1400 range: an average centipawn loss under 30 is solid, careful play. Between 30 and 60 is normal — most games at this level land here, with a handful of small slips. Above 80 usually means at least one real blunder dragged the average up. Don't chase a perfect 0; even titled players average well above that. The number matters far less in isolation than it does as a trend — if your average is steadily dropping over your last twenty games, you're genuinely improving, regardless of your win rate that week. It's also worth comparing your loss in the opening, middlegame, and endgame separately — a player who is sharp early but bleeds centipawns in time trouble has a very different problem to fix than one who blunders out of the opening.
Using it to actually improve
The score itself isn't the point — what you do with it is. After a game, don't just look at the final average; find the one or two moves that contributed the most centipawn loss and study those specifically. That single hanging piece or missed tactic is worth more of your attention than the twenty moves you played accurately. Track your average over time rather than per game, and pair the number with an explanation of what actually went wrong — a raw centipawn count tells you that you blundered, but a good coaching note tells you whatto look for next time so it doesn't happen again.
Curious what your average centipawn loss actually is? Import your last game and get a full move-by-move breakdown, free.
Analyze your game free