What the Evaluation Number Actually Means
An engine's score is measured in pawns, but it is not counting pawns. What the number is, what counts as an advantage, and why anything under 0.3 is noise.
· 3 min read
An engine evaluation is measured in pawns, and that is the source of most of the confusion about it. The number looks like a count of material. It is not.
+0.80 does not mean White is 0.8 pawns up. It means the whole position, weighing material against king safety, piece activity, pawn structure and space, is worth roughly what an extra pawn would be worth in an otherwise equal position. That is a comparison, not an inventory.
The scale
Positive favours White, negative favours Black, always. The engine does not flip the sign when it is Black's turn. If you set up a position where Black is crushing and the number reads −3.40, that is correct and you have not made a mistake.
| Reading | What it means in practice |
|---|---|
| 0.00 to ±0.30 | The engine cannot separate the sides. Treat as equal. |
| ±0.30 to ±0.80 | A small pull. Real, but one inaccuracy erases it. |
| ±0.80 to ±1.50 | A clear advantage. About a healthy extra pawn. |
| ±1.50 to ±3.00 | Large. Winning against strong opposition with accurate play. |
| beyond ±3.00 | Decisive, barring a blunder. |
M5 |
Forced mate in five moves, whatever the material says. |
The mate notation replaces the number entirely, because mate is not worth some quantity of pawns. It ends the game. A position where you are a rook down but mating in three is M3, not −5.00.
Under 0.30 is noise
This is the most useful thing on this page. A great many people look at +0.24 and conclude White is better. The engine is not saying that. It is saying it searched, found nothing decisive, and the small residue is an artefact of where the search happened to stop.
Two things follow. Small differences between candidate moves are not a ranking worth trusting: if the engine's top choice is +0.31 and its second is +0.27, those moves are equal and you should pick on grounds the engine cannot see, such as which one you understand. And a position that hovers between +0.1 and +0.3 for twenty moves is a draw being played out, not an advantage being converted.
The number moves for reasons
Watching the evaluation as you play moves is more informative than reading it once. A drop from +0.6 to −1.2 after a move means that move lost about 1.8 pawns of value, and the line the engine gives afterwards is the refutation. That is the whole basis of blunder detection in game review.
What the number cannot tell you is why. It is a single scalar summarising several competing factors, and the same +0.9 can mean a passed pawn on the sixth rank, an exposed enemy king, or a bishop pair in an open position. To find out which, play out the continuation the engine gives. The reason is always somewhere in that line, usually within four or five moves.
Where the number misleads
Engines evaluate fortresses badly, and this is a genuine limitation rather than a quirk. A position where one side is materially far ahead but has no way through a wall of pawns and a well-placed king will often read as a large advantage right up until the search is deep enough to see that nothing progresses. Endgame tablebases fix this for small numbers of pieces; nothing fixes it in the middlegame.
The other case worth knowing is the horizon effect: the engine pushes a bad consequence beyond the end of its search by throwing material at it, and reports the resulting position as better than it is. Deeper searches shrink the problem without eliminating it.
Both cases share a signature. If the evaluation is stable across increasing depth, it is reliable. If it moves substantially each time you search deeper, the engine has not resolved the position and the current number is provisional.
Common questions
- What does +1.0 mean in chess?
- It means the engine judges the position to be worth about as much as an extra pawn to White. It does not mean White is a pawn up. A material-equal position with a safer king and better pieces can read +1.0, and a position where White really is a pawn up can read 0.00 if the pawn cost too much.
- Is a +0.3 advantage worth anything?
- Not in practice. Below about 0.3 the engine is telling you it cannot separate the two sides. Between club players a 0.3 edge disappears within a few moves. It is worth noticing only when it persists across many moves and grows.
- Why does the evaluation say 0.00 in a position that looks winning?
- Usually a fortress or a forced repetition. The material or activity looks decisive, but no path exists through the opponent's defence, and the engine has found the line where nothing progresses. This is one of the cases where the number is more trustworthy than the appearance.
Try it on your own position
The analysis board runs the same engine in your browser, free. The app adds what a web page cannot: reading a position straight from a photograph, saved boards, and analysis with no connection.
