Author: Peter Berger
Date: 12:34:03 06/24/03
Go up one level in this thread
On June 24, 2003 at 11:40:18, Richard Pijl wrote: >On June 23, 2003 at 19:08:28, Jorge Pichard wrote: > >>For this game I used an Athlon 1.2 GHz for FRC_TheBaron and a Celeron 433 MHz >>for Fritz 8.0.0.8. After FRC_TheBaron castled in move 14.0-0-0 Queen side Fritz8 >>did NOT wanted to continued, therefore I set Up the position from that move on >>to allow Fritz8 to continue the game, and the game was won on the endgame stage. >>Therefore, the Castling of FRC_TheBaron_101 did NOT affected the outcome of what >>I expected from the stronger program using an inferior processor and with the >>black pieces. > >I'm not surprised. > >Fritz still is a lot stronger than the Baron. The difference in HW doesn't >really make up for the difference in strength, also in 'normal' games. > >The Baron FRC doesn't have special knowledge in evaluation for FRC games. The >only FRC addition is the castling rules. There are a lot of possibilities to >improve here, e.g: > >- The loss of castling queenside should probably be punished different when the >king is on b1 then when it is on g1. Also the initial position of the bishops >and queens (needing advancing of the pawns before they can develop) may have an >influence on this. >- Hardwired squares related to e.g. trapped bishops (on a7, h7) trapped rooks >(when the king moves to f1), piece development (no knight to c3 if there are >pawns on c2 and d4, and not on e4) >- Piece square tables with penalties for minor pieces (and some pawns) on their >original squares. > >So there is a lot of widespread knowledge in normal chess engines that do not >apply in FRC, or should be defined differently (more generic). When engines are >better adapted to FRC (perhaps the authors of Betsy and Chispa already took some >measures) I expect the gap with the 'normal' chess engines to close, especially >in the really weird initial positions with all major pieces on one side, e.g. >NNBBRQKR. Queenside castling might be very surprising here ! > >bye, >Richard. I thought about a test that might give an upper-bound for potential differences. We have one FRC position where we know that engines have "good" knowledge, the one of chess. It is also clearly one of those were castling is an important move as the central pawns have to move to allow development of the minor pieces and the king is not supposed to be save in the centre of the board. So we can have two "Nunn-like" positions namely the original starting position with one side not allowed to castle anymore. Now you could do a gauntlet with these two positions with an engine playing the non-castle side against some others and compair them to the results when castling is allowed ( but opening book not). The engine that can't castle is probably even slightly more handicapped because the engine knowledge includes occupying the centre with pawns ( which loosens its king position). What ELO difference to expect? If it's clearly less than 100 points ?, maybe 50 points ? maybe even less? - it could give a good hint on potential future development in computer FRC. I have no idea what result to expect, but my feeling is that it will be at least less than 100 points difference in strength. Peter
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.