How to Read a Chess Engine Evaluation
On this page
A chess evaluation summarizes an engine search of one exact position. Read the sign, perspective, depth, and line together.
Start with the score perspective
CalculatorGeek uses White perspective: positive favors White and negative favors Black. This convention does not switch when the side to move changes.
Centipawns and practical meaning
A pawn is approximately 1.00 in the display, but +1.00 is not a guaranteed extra pawn or win. King safety, initiative, fortress chances, and search horizon matter.
Mate scores
A mate score means the engine found a forced mating sequence within its search. Use the principal variation to inspect the defense rather than treating the label alone as proof.
Depth and MultiPV
Depth measures completed search plies, not a fixed level of accuracy. MultiPV asks the engine to retain several candidate moves and can reduce the effort devoted to the top line.
Worked evaluation examples
Suppose a position shows +0.25 at depth 12. That is approximately balanced and does not mean White is winning a quarter of a pawn. If the same position reaches +0.60 at depth 18, record the later depth and line instead of averaging the two numbers. If the display changes to M3, inspect the variation: the engine is reporting a forced mate in its current search, not a material score of three pawns.
For a Black-to-move position, a score of -1.10 still favors Black because CalculatorGeek keeps the White-perspective sign convention. This fixed convention makes two positions comparable without mentally switching the sign.
Common interpretation mistakes
- Reading a positive score as favorable to the side to move.
- Comparing two scores without checking that the FEN, engine version, depth and MultiPV setting match.
- Treating a shallow tactical line as proof of a forced result.
- Ignoring the opponent's best response in the principal variation.
- Using an engine during a live rated game instead of for study or post-game review.
Analyze the exact position
Open the Chess Analysis Board Calculator, verify the FEN, choose a search preset, and compare the returned lines and WDL model. Select a line to replay it on the board. If the position came from a game record, first choose the correct format with the FEN vs PGN guide.
Frequently asked questions
Why can two engines give different scores?
Engine version, search depth, hardware, hash state, MultiPV settings and evaluation models can change the result. Compare like with like.
Is +1.00 a guaranteed win?
No. It is an engine estimate roughly expressed in pawn units. Practical winning chances depend on the position and players.
Why compare several lines?
Alternative lines show whether the first move is uniquely strong or one of several close choices. Quick shows two lines, Standard three, and Deep four.
Sources
Use this resource
Keep exploring