Author: Larry Griffiths
Date: 20:41:33 09/22/00
Go up one level in this thread
On September 22, 2000 at 21:01:49, Mike Adams wrote: > I have done some modifications in pulsars mate detection which seems to >help. to early to see how it affects icc ratings. Pulsar will not try to get >out of check necesarily in qsearch but it will return checkmate now rather than >being allowed to pass on the move when in checkmate in qsearch. With this >improvement pulsar avoides the mate with a depth 4 search without nullmove and >depth 5 with null move. This is a lot better than depth 7 it was before. > The position actually was a rook sack not a queen sack. The queen at issue >was pulsar's meaning that pulsar could delay the mate by 2 moves if it gave up >its queen otherwise it was mate in 3 after pulsar's bad move. But your point >probably applies just as well since a rook sack would probably have a similar >affect to a queen sack on null move. I think one additional modification that >would help is that if pulsar searches depth 4 and all is dandy then it searches >3 moves in depth 5 and gets the abort search command but all 3 moves on depht 5 >return checkmate for pulsar, pulsar should continue to search the current depth >and use more than its alloted time in the hopes of finding its way out of the >checkmate. I noticed that depth 5 was a very slow search since its move >ordering at the root was invalidated once it realized that its depht 1 2 3 and 4 >search results were no good. So it may help to have more time for such >dangerous positions. Sounds like you are least having some luck with null move. I tried to implement a NULL MOVE today and got up to 50% decrease in time on one position, and another position took about 5 times longer. Larry.
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.