For most of chess history, a move got a grade only if someone wrote about the game. Annotators used six marks. "!!" for an outstanding move, "!" for a good one, "!?" for interesting but unclear, "?!" for dubious, "?" for a mistake and "??" for a blunder. When a 13-year-old Bobby Fischer sacrificed his queen in the "Game of the Century," annotators gave the move two exclamation points.

The marks were opinions. A summary of the conventions notes that "reasons for awarding the symbol vary greatly between annotators." They became an international shorthand, spread by publications such as Chess Informant, but they were never a measurement. Two experts could look at the same move and disagree.

Engines turned the grade into a number. The key idea on both big sites is the same. Instead of asking how many pawns a move gave away, they ask how much it changed your chances of winning.

Lichess converts the engine's evaluation into a winning percentage and flags moves that lower it as inaccuracies, mistakes or blunders. It does not hand out praise. Its own explanation is blunt: "good moves don't exist! You can't increase your winning chances by playing a move, only reduce them if you make a mistake."

Chess.com keeps the praise and measures with what it calls Expected Points. A score of 1.00 means you are certain to win, 0.00 means certain to lose, and 0.50 is even. After each move, it checks how much that number dropped:

Best: no drop at all.

Excellent: up to 0.02.

Good: 0.02 to 0.05.

Inaccuracy: 0.05 to 0.10.

Mistake: 0.10 to 0.20.

Blunder: 0.20 or more.

Say you have a small edge and your expected points are 0.60. You play a solid move and you are at 0.57. That is Good. You drift to 0.53, and that is an Inaccuracy. Drop to 0.45 and you have made a Mistake. Fall to 0.35 and it is a Blunder, because you have given away a quarter of a game in one move.

This is also why the labels sometimes feel strange. In a completely winning position, your expected points are already close to 1.00, so even a sloppy move may barely change the number and still be called Good. In a tense, level position, a small slip can cost a lot.

Your rating matters too. Chess.com calculates expected points from "their rating and the engine evaluation." The idea is that the same advantage is worth more to a player who knows how to convert it. So the same move can earn a different label at 900 than at 2200.

Four labels sit outside the scale. Book is "A conventional opening move." Miss is "when you fail to capitalize on your opponent's mistake." Great is for moves "critical to the outcome of the game," such as "finding the only good move in a position." Brilliant has the simplest rule of all: "a Brilliant move is when you find a good piece sacrifice." Chess.com says it is "more generous" with Great and Brilliant for newer players.

Because the main labels are ranges, several moves in one position can share a label. Three different moves might all be Excellent. Best is the exception, since it means losing nothing.