Author: Uri Blass
Date: 13:01:19 07/05/02
Go up one level in this thread
On July 05, 2002 at 15:45:21, Sune Fischer wrote: >On July 05, 2002 at 13:29:16, Uri Blass wrote: >> >>Note that a program does not have to find the long mate in order to play perfect >>chess. >> >>It only needs not to do mistakes. >> >>Let assume for the discussion that not doing mistake in KRB vs KBN position may >>be a hard problem that cannot be solved by searching 50 plies forward(I am not >>sure if it is the case) >> >>I still believe that black can avoid the trouble of being the weaker side in KRB >>vs KBN position by searching 50 plies forward so it does not prove that >>searching 50 plies forward is not enough never to lose games. > >Assume we didn't have the KRB-KBN table and that we were having this duscussion >the day before it was completed. We would be betting on a position X, and your >argument would be (if I understand you right) that because 99.5% of all the >games that reached position X ended draw, the position would be more likely to >be drawn. The table base would then show us, that white can in fact win it. > >I believe these endgames are usually easier to draw than to win. >Sometimes the winning side must play almost perfect for 100 moves, where as the >drawing side often has more than 1 alternative. Ocasionally the right winning >move has to be found by a search to very end (forced mate), the point is you can >not know if/when there are any of these moves just bcause you always get draws. > >Your example with KRB-KBN requires that kind of depth to win, the right moves >are too obscure to be found by an "evaluation" search. > >-S. I suspect that search may see that the right move help to push the opponent king closer to the corner relative to the wrong moves and it may be enough. Uri
This page took 0.03 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.