What Does +1.5 Mean? How to Read an Engine Evaluation
The evaluation bar says +1.5. The engine "thinks White is better", but by how much, and does it matter? A number like that is easy to misread. Here is what it measures, what it promises, and how to read it without fooling yourself.
What the number is
An engine scores a position in centipawns: hundredths of a pawn. +100 is worth about one pawn in White's favour, −250 is worth two and a half pawns in Black's favour, and 0.00 means the engine sees the sides as equal. Most sites display it in pawns (+1.5) and from White's point of view, so a negative number means Black is better.
The score is not a count of material. It is the engine's estimate of the position after best play by both sides, taking into account everything it can see: material, but also king safety, piece activity, pawn structure and concrete tactics it has calculated. A knight is worth about three pawns on average, yet the engine will happily rate a position where a side has given up a knight as better for that side if the compensation is there.
A rough scale
The table shows how the size of the score maps to how you should think about it, together with the winning chances the Lichess model (also used by RookSac) assigns. The model comes from games between strong players, so at club level the same score is usually a weaker promise than it looks.
| Score (pawns) | Win chance for the better side | In words |
|---|---|---|
| 0.0 to 0.3 Equal | 50% to 53% | Either side can play for a win; the result depends on who plays better. |
| 0.3 to 0.8 Slight edge | 53% to 57% | A small, real advantage. Easily lost by one inaccurate move. |
| 0.8 to 1.5 Clear edge | 57% to 63% | A comfortable advantage that still needs converting. |
| 1.5 to 3.0 Large advantage | 63% to 75% | The better side should win with reasonable play, but against a determined opponent it takes work. |
| 3.0 or more Winning | 75% or more | A decisive advantage, often a whole piece or a mating attack. Conversion is mostly technique. |
The curve is the important part. Going from 0.0 to +1.0 lifts the winning chances by about 9 points. Going from +7.0 to +8.0 lifts them by about 2. The first pawn changes a lot; later pawns change very little. That is why a review tool, like RookSac, measures the change in winning chances caused by a move rather than the raw difference in pawns (see what accuracy means).
Material is not the whole story
This position comes from the Italian Game, after 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 d5 5.exd5 Nxd5 6.Nxf7 Kxf7. White has given up a knight for a pawn, so on the scoreboard White is two points down. Stockfish, however, rates the position at about +1.0 for White.
A human would say "White is down a piece", and would be right about the material and wrong about the position. The engine is looking at what happens next: the king in the open, the threats against it, and the lack of a safe square. This is why an evaluation can't be replaced by counting pieces, and also why, at club level, a big number from a sacrifice is only useful if you can find the follow-up.
Mate scores
When the engine sees a forced checkmate it shows #5 (mate in 5) instead of a score. A mate score means the engine has found a sequence that forces mate whatever the opponent does. It is not "very high"; it is a different kind of statement. If the mate is against you, the number shows how many moves you have left. Most review tools treat a swing from a normal score to a mate score as a decisive change in winning chances, and they should.
Sharp and quiet positions
The same +1.5 can mean two very different things. In a quiet position with plenty of time it means you have a lasting positional edge. In a sharp middlegame it may be true only if you find a series of only-moves; one slip and the score vanishes. When you read a number, ask how hard is the path to make it real? The best way to find out is to play the engine's line for a few moves and see whether the required moves are natural or surprising.
Swings matter more than levels
If you remember one thing: look at how the evaluation changes after each move, not at where it stands. A position at +6 does not need studying. A position that fell from +0.3 to −1.8 after your move needs a look. In winning-chance terms that move took White from 53% to 34%, a drop of 19 points, which RookSac would label a Mistake (see move classifications).
Why the numbers move
- Depth. The deeper the engine searches, the more it sees. A quiet-looking move may turn out to lose to a tactic eight moves deep. Differences of a few tenths of a pawn between depths are normal; treat them as noise.
- Different engines. Two Stockfish versions, or Stockfish and another engine, can disagree by half a pawn on the same position. Two websites showing different bars for the same game is normal.
- Endgame tablebases. Some sites, Lichess for one, can use tablebases, which give exact results for positions with few pieces. RookSac's browser engine does not, so a theoretical draw in a long fortress may show a small positive or negative score instead of 0.00.
- Ties. Two moves scored +0.20 and +0.15 are effectively the same move. RookSac groups very small differences rather than ranking them, because the engine's finer ordering is below what any player could use.
A practical way to use evaluations
- Ignore anything beyond about ±3 unless you are studying how to convert. The game was decided earlier.
- Look for the biggest drops in a game and study those moves. See how to review your games.
- Read the engine's line to learn why the number changed, not just what it is.
- Don't chase tenths of a pawn. A move that scores 0.2 lower than the best is not a mistake.
- Keep human feasibility in mind. A move that is best only by the engine's calculation of twelve precise moves is not a lesson for your next game.