Author: Dan Honeycutt
Date: 14:54:12 06/13/05
Go up one level in this thread
Hi Chris I've had a chance to look at positions 2 and 3. You seem to say, and I concur, that Bruja handles 2 correctly. You are correct that Bruja will allow position 3. Are you sure it is illegal? Say I give up my 2 bishops, 2 rooks and queen judiciously so 5 of your pawns break through the chain. Those 5 march home as do 5 of mine on the vacated files. We still have 3 pawns each blocked. If you help out with your bishops, rooks and queen, can we not get the rest of them home? What bothers me about position 3 is the 1 move pv. This is Bruja's behavior when every move leads to a forced draw. Strange as your position is, Bruja should search it normally - ie with the number of moves in the pv roughly equal to the iteration depth. I'll have to dig into this more deeply. Bruja's maximum search depth is 63. Bruja does not send anything to the gui for the first few plys, but you can see the depth is as indicated by the length of the pv. In case you are worried about similarities to Crafty, I'll explain where that comes from. My depth based search arrays are 64 in size; ply 0 for me is the root and I can go to 63 before I overrun my arrays. I virtually always dimension arrays to be divisible by a power of 2 and I prefer a power of two proper. I do this to help the compiler with alignment of the variables. A dimension of 32 would not cut it, Bruja will go that deep in positions with many transpositions like Fine #70 and even in some normal positions if it is extending heavily and there are a lot of qsearch exchanges. A dimension of 128 would be rather pretentious on my part. Hence 64. (Of course, if you're really worried about similarities to Crafty just say the word and I'll send you the source.) I appreciate your tests. If you find other positions where Bruja does not behave as you think it should, send them my way. As I said, I'm a glutton for punishment. Best Dan H.
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.