Average centipawn loss

For each move, centipawn loss measures how much the position worsened relative to the engine's preferred option. Average centipawn loss, abbreviated ACPL, adds those values and divides the total by the number of analyzed moves. A lower value means that, on average, the player's decisions stayed closer to the choices preferred by the engine used as the reference.

ACPL is not an official score, an accuracy percentage, or a direct measurement of playing strength. Two players can reach the same average in very different ways. One may make small inaccuracies repeatedly, while another plays almost perfectly except for one very large mistake.

The character of the game also matters. Sequences in which almost every reply is compulsory offer few difficult choices and tend to produce low losses. Complicated positions may create larger gaps. A long conversion of an already winning position can add many easy moves and lower the average, while one critical mistake may dominate a short game.

Meaningful comparisons require the same engine, version, search limits, and calculation rules. Even then, the distribution of errors and the actual positions should be inspected. ACPL summarizes distance from an engine reference, but it does not explain practical difficulty, decision quality, or the reason for a mistake.