Author: Jan K.
Date: 17:56:28 09/05/04
Go up one level in this thread
>Having problems getting above the 79%-80% mark. >Been stuck there for a week or two. Not many ideas >lately. Up from 65% result in June when it was 2-3 >months old, before ICDers got a hold of it. Not sure >where to squeeze out additional advances. > >Since I don't analyze any specific position, excepting >WAC 141 going forward is not as easy. I know that WAC 2 >could be rewarded with connected passed pawns on 6th >rank evaluation term being worth a high quantity. Perhaps >I will just analyze the failures now and can gain >insight. Just take those positions one after another...first make sure that you have implemented the alfa/beta algorithm right, then make sure extensions are working correctly(check,single reply,threat,pawnpush help here much,you can try fractional extensions)...next step of course is to find if you don't solve the position at all or if it just takes too much time. I guess that there are maybe 10 positions that could take more than 5 seconds to solve. Try to turn off various parts of your search to see if it doesn't help,hash tables, see pruning,....null move may cause trouble if you prune too much->try generating quite checks at first x ply of qsearch,and so on and so on. Of course you must be absolutely sure that your "core" routines like checks, move generator and so on are working perfectly. Doing this you _must_ come finally to some logical conclusions and guess where the bug probably is. With wac positions which are simple there's usually no need to spend hours exploring the search tree, finding what happened, debugging can be made in much more effective(and clever) way.
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.