i have just aborted the match to try the new v7 build, but Iggy had taken his first lead in the match, +4 =3 -3 by winning a spanish that rybka only managed to draw. i have not observed the games and will gladly post as the match is ongoing/when its completed if there is interest. and i can figure out how to find them.
i got +80 elo in blitz in limited testing so i am curious how that changes at tournament TC. i do feel that the noomen test suite might inhibit Iggy a bit as it seemed to gain some slight advantages in the opening (i was using a heavily truncated book).
so far, just 3 more draws to report. i have decided to stick with v5fix of igorrit until a more advanced proven, stable version comes out. i hope to post the full games if i can figure out how. i think it might be enlightening to some.
i suppose it is rather pointless (to many) to say but unlike while watching the blitz games between the two, this does feel very much like watching two versions of the same engine play each other so far...(that i've hardly seen any of the games)
yanquis1972 wrote:Igorrit now +2 on the match after outplaying rybka in caro late middlegame
I am running a testmatch Nunn2 (=50 games) at 25 min each on Centrino Dual Core T4300 at 2100 Mhz.
standings sofar: Igorrit 0.086v7 2CPU - Rybka 3 2CPU : +9 =9 -6
Are you guys testing with Rybka 3 with contempt 0? Otherwise all you are doing is seeing how contempt affects the game. Make sure in your testing that you use the core value of Rybka 3 which is contempt 0. Contempt 15 was done to take advantage of weaker engines ... I have tried matches between R3 contempt 0 and contempt 15 and contempt 15 makes the engine refuse draws in totally drawn positions and will make it try to win even though its analysis evaluation says it is drawn. This pollutes the results especially if you go back and analyse the games with R3 since analysis mode is contempt 0.
beram wrote:I have tested Rybka against R..g3 with contempt = 0 and Rybka performed less. So my last tests are with contempt = 15, which is Rybkas standard value.
The last tests I have done with contempt 0 result in more losses for Rybka so I just use default settings. I am open minded about this...I will try again.
i am using contempt 15, but this was actually mainly because a)it's default b)paul's testing (abbreviated unfortunately) of long TC matches b/w rybka & robbo suggested contempt 15 is fine. also in my own analysis i sometimes find that turning contempt up actually helps in terms of finding moves. i think it may hurt the accuracy of the evaluation more than the actual engine play.
Igorrit now +3 overall after picking up another victory overnight (discarding the loss on time by v7 when it thought it had a superior position; i believe i repeated that game with v5 and it was a draw)
igorrit draws more blood. now +4 on the match...i think this game is especially interesting as while i have not looked at it closely rybka seemed to appropriately give herself a decent edge in the early middlegame, where black has much better pieces but at cost of a pawn. i am not sure how rybka lost this game. i hope M ANSARI or someone will take a look at it.
