By turbojuice1122 Date 2010-05-27 15:04
I just had Rybka 4 hang while doing a deep position analysis in Fritz 10 GUI. I have never had this happen before with an engine. It was eventually solved by right-clicking and "Properties" and "OK", forcing it to move. It was a normal, quiet position without any fail low or fail high worries. Has anyone else had this happen with Rybka 4? To whom should I e-mail a screen shot and more details?
By Kapaun Date 2010-05-27 15:36
Yes, I have, and I mentioned it to Vas, but he just said it could happen that she uses extremely much time for one move. I have had this also at times with R3, by the way. I guess in your case it was a move which was for some reason unclear for her.
By turbojuice1122 Date 2010-05-27 20:06
No, I can say with 100% certainty that this is definitely NOT the case (Vas's explanation). This is actually a major bug. These are moves that take a few seconds suddenly taking a few hours. There is nothing like it having a fail high or fail low due to finding that a move is much better or worse. It also only happens when it is using MPV.
By Kapaun Date 2010-05-27 20:10
Well, tell ist Vas, my friend...
By Vempele Date 2010-05-27 20:21
Try setting "Always Score Main Move" to true.
By turbojuice1122 Date 2010-05-27 20:27
Ahh, very interesting--now it works! Why is this? Thank you. :-)
By turbojuice1122 Date 2010-05-27 20:32
By the way, when this was occurring, before I tried your solution, I noticed that on at least one occasion, the nps count more than doubled compared with what it was for the other moves in these similar positions--I was supposedly getting 233 kn/s on a 32-bit dual-core laptop.
By Uly Date 2010-05-27 20:39
> There is nothing like it having a fail high or fail low due to finding that a move is much better or worse.
If "Always Score Main Move" solved it, then it was related to fail lows Rybka was taking ages to resolve (but didn't show anything to the user.)
By turbojuice1122 Date 2010-05-27 20:52
That would be very strange, since all of these "fail lows" were draw scores--scores of like 0.06 changing to 0.00, etc.
By turbojuice1122 Date 2010-05-27 20:53
Also, it doesn't take three hours to resolve a "fail low" at depth 12 in an endgame. :-)
By turbojuice1122 Date 2010-05-28 15:39
Whoops, I was wrong--this only "acted" like it fixed the problem, but the problem is still present--at least when I "Create UCI Engine" specifically for DPA with this box "checked".
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill