Author: Bas Hamstra
Date: 05:51:32 09/26/00
Go up one level in this thread
On September 26, 2000 at 07:06:07, David Rasmussen wrote: >On September 25, 2000 at 13:47:03, Robert Hyatt wrote: > >>On September 25, 2000 at 11:41:18, David Rasmussen wrote: >> >>>Am I the only one who >>> >>>a) thinks that the kind of null move pruning that is done in Hyatt's Crafty is >>>the reverse of what Ernst Heinz concludes in his paper about ANMP >> >> >>I hope not. IE here is my code for that: >> >> null_depth=(depth > 6*INCPLY) ? 4*INCPLY : 3*INCPLY; >> >>Which says if the remaining depth is more than 6 plies, use R=3, while >>if the remaining depth is <= 6 plies, use R=2. >> > >Of course, now I see. My brain must have been lacking oxygene or something... > >>> >>>I.E. When I do >>> if (depth>6) >>> R=2; >>> else >>> R=3; >>>I get the best results. >>>Why? >> >> >>What is depth for you? Current ply? Or plies remaining? For me it is >>plies remaining before dropping into q-search, which means I am doing the >>same thing as Ernst, basically. Your code is the exact opposite of what we >>are doing, assuming 'depth' means plies remaining and not depth from root of >>the tree. > >I do the same as you and Ernst, as far as depth is concerned. "depth" means >remaining ply in my program. And I've done extensive testing on both tactical, >quiet and endgame positions, and in all three cases, my "reverse" (compared to >yours) concept is by far the best for my program. It saves about 25% nodes on >all test positions put together. > >Why is that... I don't get it. Have you tested on enough positions? For instance all 300 wac? And, given the same time, is the same number of solutions found? And if all that looks still good, there's how it does in real games, does it find the subtle moves too? It might be so that your system only works well for "hacking wood". For me R=2 seems pretty safe overall. Personally I have the idea that R=3 destabelizes the search. Some positions it searches deeper, and some it searches even less deep. And some positions is seaches deeper at the cost of overlooking tactics. Regards, Bas Hamstra.
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.