Author: scott farrell
Date: 18:51:56 01/14/03
Go up one level in this thread
On January 14, 2003 at 20:38:59, Dann Corbit wrote: >On January 14, 2003 at 19:44:44, scott farrell wrote: >>On January 14, 2003 at 16:28:45, Dann Corbit wrote: >>Dan, >>Firstly thanx for your interest on this. >>No it doesnt solve it. >> >>Again its time to ply is slower, it get 1 ply less with the same time approx. >> >>Interestingly though, with the blockade code turned on when material<Queen, ie. >>all positions in wac230, the score starts at +5 (mainly because of the 2 linked >>passers +1 pawn up), but the score goes does each ply, failing low on most, so >>after a few plies the score is down to +3. >> >>I think this is good, it didnt give a straight out draw, which it is not, with >>the fiddling sort of move the rook around and defend etc, the score starts to >>tend towards a draw. Given their are possibilities of moving pawns forward for >>both sides, I think this is as good as I can expect from my code a this point. >> >>My chompster will score the move c3 after the sac as about an extra +4 to the >>score, so immediately it wont think it got return on its sac, probably if I >>pushed the eval of a linked passers higher, it would sac more quickly. > >That is one way that it can easily be solved. However, it makes the programs >play like crap. If you run at "good playing strength" values for passed pawns, >it is much harder to find. Only crafty and ruffian solve it, that I remember. > Oh yeh, I agree, I didnt say i was going to change it, just was one way to solve it. It doesnt think the whole position is blocked, it just knows not to make it more blocked. >You could (perhaps) add to your algorithm a heuristic like this: > >1. I am ahead on material >2. I detect a blockade >3. Boost passed pawns >4. Set value of breaking the blockade to my material advantage less 1 pawn. > yeh I like the ideas above. I will have to think some more. Maybe if during say during search of ply8 I detect x number or x% of blockades, for ply9 I can adjust some values for passed pawns etc. But I think I run into trouble with the hash table having older values, I need to think on it more. >Some programs detect the problem (e.g. Hossa) but they don't know what to do >about it, in order to solve it. > >>>How does your algorithm do on WAC.230? >>>[D]2b5/1r6/2kBp1p1/p2pP1P1/2pP4/1pP3K1/1R3P2/8 b - - bm Rb4; id "WAC.230";
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.