What the number next to the board actually means, and how to use it without over-trusting it.
When Stockfish evaluates a position, it outputs a number in pawns — roughly, how much material advantage White has, according to the engine. +1.0 means White is about a pawn ahead. -1.0 means Black is a pawn ahead. 0.00 means the engine sees the position as balanced. It isn't literally counting pawns on the board — the engine folds in piece activity, king safety, and pawn structure too — but "pawns of advantage" is the right mental model.
Sometimes you'll see something like #3 instead of a normal number. That means forced checkmate in 3 moves, not a material count — mate scores always override the regular evaluation because nothing about material matters once mate is forced. A #-2 means the other side has forced mate in 2. See The Mate treats mate as an extreme value internally specifically so it never gets confused with a merely large material advantage.
On the Analyze page, the evaluation graph doesn't plot the raw number in a straight line — it's compressed with a curve. Here's why: the difference between +1 and +2 is meaningful (a small edge becoming a real one), but the difference between +8 and +12 barely matters — both are completely winning. A straight-line graph would waste most of its height on already-decided positions and squash the genuinely close, interesting part of the game into a flat line near zero. The curve keeps close games readable while still showing when a position becomes clearly winning.
In practice, evaluations under about ±1.5 are still a real fight — either side has practical chances, and a human opponent (unlike Stockfish) can absolutely go wrong from there. Evaluations past about ±3 are usually decisive with correct play, though "decisive with correct play" and "actually winning against a real opponent" are two different things — this is why a move that keeps a +2 position at +2 gets rewarded, even though the engine had a flashier line that also stayed winning.
The evaluation number and the move labels (Best, Inaccuracy, Blunder, and so on) are two views of the same underlying data. The label for each move is based on centipawn loss — how much the evaluation dropped compared to the engine's own best move at that position, from the perspective of whoever just moved. A full breakdown of every label and its exact threshold is in the move classifications guide.
Don't chase 0.00. A position doesn't need to be perfectly equal to be fine for you — plenty of strong, well-known positions sit at +0.3 or +0.6 and are simply normal, healthy chess. What actually matters when reviewing your own games is the direction and size of swings move to move, not whether any single number is exactly zero.
Ready to see this in action on a real game? Head to Analyze and paste in a game, or try it live from any position on Explore.