The evaluation bar is the strip beside the board that fills with white or black as an engine thinks. It is the most-watched part of any analysis screen and one of the most misread. This guide explains what the number means and how to use it without being misled.
The basic idea
An engine searches many moves ahead and returns a single score for the position. By convention:
- a positive score means White is better,
- a negative score means Black is better,
- 0.00 means the position is equal or a forced draw.
The bar is just a picture of that number. When it is mostly white, White is better; when it fills with black, Black is better. Most bars are scaled so that small advantages move it a little and big advantages move it a lot, with a cap so a winning position does not push it off the edge.
Centipawns and what +1.0 means
The traditional unit is the centipawn, one hundredth of a pawn. A score of +100 centipawns is shown as +1.00 and means, roughly, "White is better by about the value of one pawn".
That does not mean White is literally a pawn up. It means the engine judges White's position to be worth about as much as an extra pawn would be in a normal position. White might be a pawn up, or have a much safer king, or a far more active rook, or some mix of these.
A rough guide for players:
- 0.00 to ±0.30: level. The opening position is usually given a small plus for White, because White moves first.
- ±0.30 to ±0.80: a slight edge. Good players draw these positions often.
- ±0.80 to ±1.50: a clear advantage. With accurate play the better side has real winning chances.
- ±1.50 to ±3.00: a large advantage, often winning in practice.
- Beyond ±3.00: usually decisive, for example a piece up with no compensation.
These bands are rules of thumb, not laws. A +1.0 in a queen-and-rook middlegame is very different from a +1.0 in a pawn endgame, where a single pawn is often the whole game.
White: bishop on b5, pawn on e4, knight on f3, pawn on a2, pawn on b2, pawn on c2, pawn on d2, pawn on f2, pawn on g2, pawn on h2, rook on a1, knight on b1, bishop on c1, queen on d1, king on e1, rook on h1. Black: rook on a8, bishop on c8, queen on d8, king on e8, bishop on f8, knight on g8, rook on h8, pawn on b7, pawn on c7, pawn on d7, pawn on f7, pawn on g7, pawn on h7, pawn on a6, knight on c6, pawn on e5. White to move.
Win probability
Some analysis tools also show the evaluation as a win chance or expected score, for example "White 62%". That number is a translation of the centipawn score into "how often does a player with this advantage score points". It is often more useful than raw centipawns, because the difference between +0.2 and +0.7 matters much more than the difference between +6.2 and +6.7. In both cases the gap is half a pawn, but only the first changes the likely result.
Mate scores
When an engine finds a forced checkmate, it stops showing centipawns and shows the number of moves to mate instead, usually written M3 or #3. A positive mate score is a mate for White; a negative one, such as -M2 or #-2, is a mate for Black.
White: pawn on f2, pawn on g2, pawn on h2, rook on d1, king on g1. Black: king on g8, pawn on f7, pawn on g7, pawn on h7. White to move.
A mate score is the most reliable thing an engine can tell you: it has seen the whole forced line to the end. The count can still shift when the engine searches deeper and finds a faster mate, but a mate it has found does not disappear.
Why the evaluation keeps changing
Beginners are often confused when the score jumps from +0.4 to +1.1 to +0.7 while the engine is still thinking. There are two different reasons for movement.
1. The engine is searching deeper. Engines look ahead in steps called depth. At a low depth the engine may not yet see a defence or a tactic, so the number can swing. As depth increases, the evaluation usually settles. If you are checking an important moment, give the engine a few seconds before trusting the number.
2. A move has changed the position. This is the movement that matters for your games. When you play a move and the evaluation drops from +1.2 to -0.5, your move gave away about 1.7 pawns of value. The game review uses exactly these drops to decide which moves were mistakes.
Different engines, versions and settings can give slightly different numbers for the same position. A gap of 0.2 or 0.3 between two tools is normal and means nothing.
When not to trust the number
An engine is far stronger than any human, but its single number hides things that matter to you.
Fortresses and drawish endgames
Some positions are a material advantage but a dead draw. Engines without endgame tables, or at low depth, can show a plus score where no win exists.
White: bishop on c3, king on f2. Black: king on f7, bishop on d5, pawn on f5. White to move.
Here Black has an extra pawn, which on material alone looks like an advantage. But the White bishop controls the dark squares and the white king can blockade the pawn. There is no way through. A strong engine will show a score near 0.00 for this, but in many similar endgames with more pawns the engine may show a small plus that does not translate into winning chances.
White: king on g6, pawn on h6, bishop on d3. Black: king on h8. White to move.
This one surprises many players. White is a bishop and a pawn up, but it is a draw: the bishop does not control the promotion square, and the black king just shuffles between h8 and g8 (or gets stalemated). Engines with endgame tables see 0.00 at once.
Positions that are hard for humans
A +0.5 that requires finding five only-moves in a row is not a +0.5 for you. A -0.5 in a simple position with clear plans may be easier to play than a +0.5 in chaos. When you choose between two moves of similar value, prefer the one whose follow-up you understand.
Short thinking time
If the engine has only searched for a fraction of a second, treat the number as a first impression. This matters most in sharp tactical positions, where one hidden move can flip the score completely.
Opening evaluations
An engine may give +0.3 to one opening and +0.1 to another. That difference is far too small to decide what you play. Choose openings by how well you understand the resulting positions, not by the second decimal place. Opening training helps you learn the plans behind the moves.
How to use the bar well
- Turn it off while you think. Look at the position on your own first. If the bar is visible, you stop calculating and start guessing what the number means.
- Use it to find moments, not answers. The graph of the evaluation over a game shows where the big swings were. Go to those moments and work out why the swing happened.
- Look at the best line, not just the number. The engine's main line explains the score. A +2.0 based on a tactic you have not seen is a lesson; a +2.0 alone is just a fact.
- Think in results. Ask "did my move change the likely result?" rather than "did I lose 0.3?".
The bar is a measuring tool. Used with your own thinking, it shows you exactly where your games are won and lost.
Published 2026-11-01 · 22Softwares