our actual rating lists are online and can be found under the attached links.
40 / 20:
New games: 1712; 28 different engines
Total: 714.592
NEW Engines 228 DiscoCheck 5.2 x64: 2769 - 848 games (+112 to v. 5.0.1) 524 Nirvana Chess 1.4 x64: 2612 - 481 games (+20 to v. 1.3)
UPDATES 2 Stockfish DD x64 4CPU: 3148 - 1052 games (-9) 396 SmarThink 1.50 x64: 2677 - 1200 games (-4) 435 DiscoCheck 5.0.1 x64: 2657 - 807 games (-8) 562 Arasan 16.2 x64 1CPU: 2587 - 800 games (+2) 524 Nirvana Chess 1.3 x64: 2592 - 900 games (+7)
40 / 4: (from Jan. 09th)
no update;
we are testing: DiscoCheck 5.2 x64 2751/500; Chiron 2.0 x64 4CPU +54/400; Vajolet 2.48 and EveAnn 1.72 - see more in our Forum.
5'+3'' pb=on
no update;
We started to test DiscoCheck 5,2 x64: It would be nice to hear from Lucas - as we have crashes too with 5 or less pieces on the board inside Shredder GUI or Arena 3.5!
Werner wrote:
5'+3'' pb=on
no update;
We started to test DiscoCheck 5,2 x64: It would be nice to hear from Lucas - as we have crashes too with 5 or less pieces on the board inside Shredder GUI or Arena 3.5!
Sorry, I forgot to answer your PM. It has also been pointed out by other people in another thread. Yes, DiscoCheck is very broken in pondering mode. It becomes especially apparent when you disable the GUI adjudication/resign. One bug I can reproduce (depth 127) and another one (not delivering mate while keeping mate in sight) I cannot reproduce deterministically although I have seen it too. It's a hash table problem: the worst kind...
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.
lucasart wrote:
Sorry, I forgot to answer your PM. It has also been pointed out by other people in another thread. Yes, DiscoCheck is very broken in pondering mode. It becomes especially apparent when you disable the GUI adjudication/resign. One bug I can reproduce (depth 127) and another one (not delivering mate while keeping mate in sight) I cannot reproduce deterministically although I have seen it too. It's a hash table problem: the worst kind...
Thanks Lucas,
the second problem causes to Crash the GUI too - so it is not easy to find out what happens. Perhaps a log can help you ?
At the moment I use CB GUI with tbs. This is the only way I can avoid these problems.
Werner wrote:
Mon Jan 20 11:15:18 2014: from DiscoCheck 5.2 x64 (1): info score mate -2 depth 127 nodes 3988 time 0 pv a2a1n c2b1 a1c2 b1b3
Mon Jan 20 11:15:18 2014: from DiscoCheck 5.2 x64 (1): bestmove a2a1n ponder c2b1
This is the bug! When iteration depth=127 finishes, the iterative deepening loop stops, DiscoCheck returns the best move and stops searching. The UCI protocol explicitly says that engines should never stop when pondering, which means some artificial code needs to be added to prevent it (ideally a blocking I/O read). I will fix this in the next version. In the meantime, we will have to live with the fact that DiscoCheck is buggy in pondering mode
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.
