Standings so far:
- Dr.Deeb 5/6
Tigran 2.4n x64 4/6
Patzer 3.80 3/6
Simplex 0.9.8 x64 0/6
Tigran lost with the black pieces to Patzer which was a surprise for me ....
Cheers
Moderators: hgm, Rebel, chrisw
Nice games! With the first one, I think you could have continued the game with 32. ... Rb8 where White has advantage, but I'm not sure whether this is already won at that level. The second one shows a major weakness of the engine when it comes to king attacks. I've tried many things with king danger zones, but so far, none of that increased the overall strength. The workaround is getting some centre control instead, but where that fails, loss is likely.Dr.Wael Deeb wrote: ↑Fri Jul 17, 2020 12:11 amAnd the chess engine that crushed the human in 2 games :
CT800 1.40 x64
Hi Ras,Ras wrote: ↑Fri Jul 17, 2020 11:16 amNice games! With the first one, I think you could have continued the game with 32. ... Rb8 where White has advantage, but I'm not sure whether this is already won at that level. The second one shows a major weakness of the engine when it comes to king attacks. I've tried many things with king danger zones, but so far, none of that increased the overall strength. The workaround is getting some centre control instead, but where that fails, loss is likely.Dr.Wael Deeb wrote: ↑Fri Jul 17, 2020 12:11 amAnd the chess engine that crushed the human in 2 games :
CT800 1.40 x64
What makes this engine a bit special for human play is that it's designed to play humans, not other engines. That's why it doesn't like to trade down for no reason since a full board invites more human tactical errors. It also doesn't like closed centres because that's a well known human strategy against computers. The small internal opening book (23k moves, 13k positions) favours variety over strength, intended to play at club level.
If you throttle the node rate to about 30 kNPS (in the UCI options) and set the hashtables to their minimum of 1 MB, the engine is very close to the actual dedicated unit with the Cortex-M4 microcontroller.