Author: Dieter Buerssner
Date: 16:08:30 05/04/02
Go up one level in this thread
On May 04, 2002 at 08:30:51, Uri Blass wrote: >Maybe the problemin Fritz is because of trying to construct the pv out of hash >tables(I have no idea how programs do it) but I think that in this case it is >better not to print the pv at least in cases that it includes repetition. I think, this is very likely. And the reasons, how this can happen, can be seen in this thread. >I know that fritz7 can display stupid pv's because of trying to construct the pv >out of hash tables and I think that it is better at least to check if the pv >makes sense before printing it. Well - in the case of repetition in the PV and score != draw_score it may be easy to detect. But this will only be a small fraction of the cases, where wrong PVs can come out of the HTs (one (the?) typical problem is, that the contents of the HT are different at search time compared to the time, the PV is displayed). Sometimes, the PV out of HTs may have valuable information (especially the earlier moves), sometimes, it may be plain stupid. In text mode, my engine tries to warn the user by an appended "H" to every move out of the HT, that this move is less reliable than other moves in the PV. Under the commercial GUIs, however this cannot be made visable. I mentioned an idea to one author of UCI - Stefan Meyer Kahlen - so that the protocol may distinguish between the more reliable and less reliable moves (for example print black and grey). I fear, (because most engines will not have that information readily available without special programming), that the interest was very limited. So the engine author practically has to decide, if he wants to give all info he has (with the danger that some might be stupid), or to hide some (possibly valuable info). I think, neither solution is good, and for me, I though to show all I have, is better. Regards, Dieter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.