Author: Mike Adams
Date: 15:22:08 10/03/00
Go up one level in this thread
Obviously if I can improve the depth of the program i will do that. Its depth has increased in recent weeks. Also this game were it only saw depth 6 for gxf3 was a blitz game were it only had a few seconds to move. Pulsar is playing about 2300 icc blitz and nearly 2500 icc bullet. This is not bad for me, i remember when pulsar was under 2000 blitz. One of the things that was irratating about this move gxf3 and the whole game that pulsar played was it was playing a 1400 player. If it had not made some dumb evaluate decisions deteriorating its postions it probably had the tactical strength to put this 1400 under but a poor evaluate makes an inconsistent program. I calculated Pulsar's performance rating that night with the 20 games in its history. I threw out the game it lost to the 1400 because of those dumb evaluate and the other 19 games it had a performance rating of 2340 so i think i'm doing something right. Adittionally i've been told that a better evaluate will increase detph. It is able to distinguish more betweeen moves getting more beta cuttoffs as it begins to see moves as obviously better or worse. But i should point out that I have played lambchop a program rated 200 points higer on the icc and which is probably even stronger if it was running on better hardware and lamchop can beat pulsar. What is interesting is lambchop gets about the same depth as pulsar when they play. So its not all deep search. But if i can increase the depth I will. right now i use hashing for ordering the first move, nothing else. I use a capture sort algorithm. I use killer moves and futility pruning. Aslo i use R2 Null Move. I do not ponder or use windowed search. I hear window search does not do much. I'd be interested in implementing pondering. Suggestions for what will produce more beta cuttoffs are welcome. I am going to work on increasing nodes per second. Some routines have been rewritten and others can be rewritten by implementing the same new ideas that i have allready begun to work on.
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.