Author: Odd Gunnar Malin
Date: 23:42:45 08/06/02
Go up one level in this thread
On August 07, 2002 at 01:27:36, Mike Byrne wrote: > > >I just created a new command "md" <move delay> that is NOT in the executable >emailed out earlier today. > >This will be standard in all future modified crafty's - if you would like the >current executable 17.16 SE with this option. Please email me at >mfbyrne_pa@msn.com. I'm not sure what you want to gain with this. Some sort of time handicap to reduse the strength? If this is true I would suggest another command for this. dd <ply> Dynamical search depth. The search depth is redused with the parameter <ply> for each search. Ex. On a move the engine reach a searchdepth at 10 plies. With a 'dd 3' it will then pick the move it found at ply 7. You have to save the best move at each ply as the search go on. To make it look nice you can also post the pv when the move at ply-dd is search for, but I guess this will not be easy with the normal triangle save of PV. With this you will get the engine to use the same time when playing at redused strength as it does with the strongest settings. You do need to keep an eye on the 'dd' parameter in relation to what rating the user want to set the engine to for setting a lower bound for the ply. For Crafty I think you can go the hole way down to 0 ply if the user use a 'dd 10' or something since it do a quence search for first move ordering if i remember right. I have added this to my (not yet released) version of PolarChess but have not had time to test it too much. I call the variable strength instead of 'dd' and add it with a command like 'set strength 3'. I use the 'set' command for all tuning of the searchengine/evaluation. In addision to this, you have to let it make some real blundermove inbetween. I have not come to this yet with my engine but my first try will be. 1. Find some common pattern when a user blunder (bishop goes backwards etc.) 2. Keep an eye on a fall low and match it with the error pattern. 3. Select the error move if depth and a probability could match the users wish of redused strength. The depth should not be constant, but varie a little. Odd Gunnar
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.