My new siteaddress

[icander]

http://privat.bahnhof.se/wb432434/welcome.htm

[Vinvin]

I'm looking one more time at the positionnal test suite page : http://privat.bahnhof.se/wb432434/fentest.htm

Do you really think the move played by Deep Rybka 3 Q6600 on pos #15 worth 0 point ?
It's a bit strange.

thanks,
Vincent

[Dann Corbit]

I'm looking one more time at the positionnal test suite page : http://privat.bahnhof.se/wb432434/fentest.htm

Do you really think the move played by Deep Rybka 3 Q6600 on pos #15 worth 0 point ?
It's a bit strange.

thanks,
Vincent

At ten minutes per position, it is possible to get wrong evaluations with even strong computer programs for positional problems.

I have a suggested experiment:

Take a very strong UCI program such as Rybka 3 and give 8 hours per position with multi-pv set to 10 on a very high end machine.
Since there are 15 positions, we can do three positions per day, for a total of 5 days crunch time.

At the end of the experiment, we will have two sets of scores -- the computer scores and the human scores.

I guess that most of them will line up, but for those that do not, we can individually analyze them to find out who is right and who is wrong.

[yanquis1972]

what move is this? on my test she seems to prefer Kf2?

--oh, i didn't see the results section. i think it must've been bad engine luck on that particular run, or else i had good luck.

i reran the position in mpv-3, & it seems rybka prefers h3 by a good margin, so that was probably the zero move. bizarre, but maybe she's onto something, who knows.

[+0.48] d=20 1.h3 (0:08:44) 252715kN
[+0.35] d=20 1.Kf2 Rxe1 2.Rxe1 Qc7 3.g3 (0:12:09) 370833kN
[+0.31] d=20 1.Re6 Nh6 2.Qh3 c4 3.Ba3 b4 4.Bc1 cxb3 5.axb3 Ng8 6.Be3 f5 7.Rxe7 Qxe7 8.Re1 Qc7 9.Re2 Qe5 10.Kf2 f4 11.Bb6 (0:12:09) 370833kN