Author: Reinhard Scharnagl
Date: 01:20:24 11/19/05
Go up one level in this thread
On November 19, 2005 at 01:56:42, Ed Trice wrote: >I am wondering if any programs incorporate the concept of fractional ply search >extensions as leaf nodes are encountered that are "near" the tablebases. For >example, in my own program, Gothic Vortex, I have the complete 5-piece tablebase >available for probing, and about 128 MB of buffers set aside for this (overkill >is my middle name.) Hi Ed, I am still in doubt, whether this would be beneficial for a searching program in general or not. Indeed there might be thinkable cases, where a detected positive ending could turn an otherwise lost or drawn game. But I am still convinced, that the distance of the found lines leading into the table range mostly will be far away from the main PV line. If there would be a clear benefit, then it should be possible to create a statistic, where e.g. following is correlated: a) the amount of change of the calculated evaluation where that approach has been used, b) the distance to the final evaluation of the main PV line. In any case there will be situations (already near to table bases) where your idea doubtlessly will produce significant improvements. Nevertheless the question is, if then the estimated outcome of the game could be changed by this at all. But there is also another aspect, which lead me still not to use huge looking up knowledge like opening libraries or table bases: even if relying on such knowlegde primarily could improve a program, in the long time run of developing a chess program this would be counter productive to test and implement changes to make the search itself more intelligent in that game stages. Because of that I think, that implementing looking up knowledge should be the very last thing to implement into a chess program, when all other ideas would have been exhausted. But I find myself very lonely here with such an idea. Thus in SMIRF I prefer variants like 8x8 Chess960 (Fischer Random) or 10x8 CRC (Capablanca Random), because of the lack of documented opening knowledge. I regard those games to be much better fitting drosophilas for to improve chess programming, because the engines are forced to survive with their own abilities alone. Best Regards, Reinhard.
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.