Author: scott farrell
Date: 21:30:40 01/01/03
I really liked Ed's premise 'reduce depth at EVERY node, unless you can provide it is not safe for some reason'. This is the opposite to me thinking before hand, and from most reading I had done, ie. razoring et al. where the idea is to prune if there is a big enough margin for safety and a lot of guess work. I liked Ed's idea (http://members.home.nl/matador/chess840.htm) of using THREAT as well as margin before doing a depth reduction, which worked nicely with his examples, a capture is good, but threatening to capture more, or threaten mate are all obviously better, so better not reduce depth there. My idea is, instead of running the expensive evaluate function, run the MORE EXPENSIVE qseach function instead of using THREAT. My qsearch finds mates involving captures, and has lots of pruning in it anyway (aka stand-pat from crafty and Robert). My idea is , whether you reduce depth or not, you are eventually going to end up do a qsearch anyway, I hash the qsearch, so that helps also. It also helps move ordering. The only time I guess you might not eventually call qsearch is when a null move causes an almighty cutoff, but as far as I know, everything else ends up in the qsearch sooner or later. I ended up getting more qsearch nodes, but got more and hopefully safer reductions. What do other's think? Hopefully I wont just get hosed by the first reader for a change. Scott
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.