How to read chess evaluation

Sort:
Jeanne1

When an engine gives a position a value of say 0.1 or 0.13 which is better? I was assuming that the higher the percent the better, however I'm looking at a game where one move rates the position as 0.23 and another engine says 0.1

Scala

Usually this mean that white have the advantage, but is so small that you can say that the positions is equal...

-0.23  black advantage

Scala

For better understand. The piece value is

Pawn 1

Knight and bishop 2,5 each

Rook 5

 

Quuen 9

HilarioFJunior

If one move is evaluated with 0.1 and another with 0.13, probably you should analyze the strategic implications of both, once tactically they are almost equal. 

Jeanne1
Scala wrote:

Usually this mean that white have the advantage, but is so small that you can say that the positions is equal...

-0.23  black advantage

So at what percent would it be clear that one side has a signifigant advantage? 1.0 and up since it's a pawn?

pilotk9

Depends which color you are

Scala
Jeanne1 wrote:
Scala wrote:

Usually this mean that white have the advantage, but is so small that you can say that the positions is equal...

-0.23  black advantage

So at what percent would it be clear that one side has a signifigant advantage? 1.0 and up since it's a pawn?

Yes... If you are white and have a pawn advantage, you should see something like ex.. 1,25 ( 1 for the pawn) if you have a knigth advantage you should see ex 3.0 (2,5 for the knight)...

leiph15
Jeanne1 wrote:
Scala wrote:

Usually this mean that white have the advantage, but is so small that you can say that the positions is equal...

-0.23  black advantage

So at what percent would it be clear that one side has a signifigant advantage? 1.0 and up since it's a pawn?

0 to 0.25 shows as = (equal)
0.26 to 0.70 shows as +/= (slight advantage)
0.71 to 1.40 shows as +/- (clear advantage)
1.41 and above shows as +- (decisive advantage)

These are approximate, just guidelines. Notwithstanding engines can be wrong, disagree with each other, and even disagree with themselves, not all positions are equal. 1.0 by itself doesn't guarantee whether a human can expect winning chances in a given position. Neither does it guarantee the position is winning with best play. Similarly 0.00 doesn't guarantee the position is drawish or will be a draw with best play.

The types of positions where you can distrust the evaluation are generally openings, closed positions, endgames and positionally based attacks on the king. In the last two you can also distrust that it's finding even adequate moves (much less best moves).

In an open middlegame position with lots of tactics an engine is very trustworthy.