Fermin Serrano wrote:I dont know if this exits: is there any epd file with positions and scores that could be classified as "the same for all engines", I mean, that evaluations would be similar to differents engines with a +- 30 or 40 centipawns?
or does anyone know of possitional positions from which I can test my eval function?
thx
I can write a hypothetical engine that emits random numbers. Therefore your question has the answer : "No."
On the other hand, most engines will emit a score of zero for draws and many will emit a scores very close to 32767 for nearby checkmates (though some use a different scale such as 99999 for checkmate).
You will find that Shredder scores higher than other engines when he thinks that he has an advantage.
For positional positions, I think the most famous one is Tony's:
http://home.interact.se/~w100107/fentest.htm
As far as matching the exact score, I would not worry about it as long as you get the right answers.
Most people calibrate their engines to that a pawn is worth 100 centipawns.