Evaluations in centipawns and symbols
Moderator: Ras
-
Marek Soszynski
- Posts: 587
- Joined: Wed May 10, 2006 7:28 pm
- Location: Birmingham, England
Evaluations in centipawns and symbols
What are the usual conversions into symbols for centipawn evaluations? I remember that +3.00 or greater is +-, but what about +/-, +/=, and = ?
Marek Soszynski
-
Andres Valverde
- Posts: 608
- Joined: Sun Feb 18, 2007 11:07 pm
- Location: Almeria. SPAIN
- Full name: Andres Valverde Toresano
Re: Evaluations in centipawns and symbols
IIRC i use the following intervals in ECTool :Marek Soszynski wrote:What are the usual conversions into symbols for centipawn evaluations? I remember that +3.00 or greater is +-, but what about +/-, +/=, and = ?
0-25 cp = (even)
25-50 cp += (white is slightly better)
50-200 cp +- (white is better)
>200 cp ++- (white is winning)
Saludos, Andres
-
Marek Soszynski
- Posts: 587
- Joined: Wed May 10, 2006 7:28 pm
- Location: Birmingham, England
Re: Evaluations in centipawns and symbols
Interesting. To say that +.50 is +/- or +- seems a very low threshold.Andres Valverde wrote:IIRC i use the following intervals in ECTool :Marek Soszynski wrote:What are the usual conversions into symbols for centipawn evaluations? I remember that +3.00 or greater is +-, but what about +/-, +/=, and = ?
0-25 cp = (even)
25-50 cp += (white is slightly better)
50-200 cp +- (white is better)
>200 cp ++- (white is winning)
Marek Soszynski
-
carldaman
- Posts: 2287
- Joined: Sat Jun 02, 2012 2:13 am
Re: Evaluations in centipawns and symbols
Many engines give +/- when above .70, but you need to know your engines...Marek Soszynski wrote:Interesting. To say that +.50 is +/- or +- seems a very low threshold.Andres Valverde wrote:IIRC i use the following intervals in ECTool :Marek Soszynski wrote:What are the usual conversions into symbols for centipawn evaluations? I remember that +3.00 or greater is +-, but what about +/-, +/=, and = ?
0-25 cp = (even)
25-50 cp += (white is slightly better)
50-200 cp +- (white is better)
>200 cp ++- (white is winning)
With Stockfish the eval would have to be well over 1.00 for me to agree with the +/-.
Regards,
CL
-
lkaufman
- Posts: 6297
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
- Full name: Larry Kaufman
Re: Evaluations in centipawns and symbols
My recollection is that the Chessbase interfaces use +.25 to +.75 for small advantage, .75 to 1.25 for clear advantage, and 1.25 or more for winning advantage. I think this is about right for engines with "normal" evals (i.e. a knight up is +3 or slightly higher), in which category I include both Houdini 3 (but not earlier Houdinis) and all Komodo versions. The evals of Stockfish (too large) and older Houdinis (too small) were incompatible with the Chessbase scale I think. Of course it's not perfect; sometimes an engine may show +2 or more in a theoretically drawn ending, especially if without tablebases.
-
Marek Soszynski
- Posts: 587
- Joined: Wed May 10, 2006 7:28 pm
- Location: Birmingham, England
Re: Evaluations in centipawns and symbols
Can't we look this up somewhere rather than rely on our recollections, or on trial and error? Are the symbol-evaluation equivalents the same for ChessBase, Aquarium and other programs?
I know that Chess Assistant can set thresholds for !! and ?! etc. in automatic analysis, but I don't know know what the defaults are.
I know that Chess Assistant can set thresholds for !! and ?! etc. in automatic analysis, but I don't know know what the defaults are.
Marek Soszynski
-
Marek Soszynski
- Posts: 587
- Joined: Wed May 10, 2006 7:28 pm
- Location: Birmingham, England
Re: Evaluations in centipawns and symbols
I found the following defaults in Scid:
Options -> Game Information -> Configure informant values
+- 3.0
+/- 1.5
+/= 0.5
http://scid.sourceforge.net/
Options -> Game Information -> Configure informant values
+- 3.0
+/- 1.5
+/= 0.5
http://scid.sourceforge.net/
Marek Soszynski