Athlon XP2800+
Fritz 10 GUI
128mb hash each
3-4-5 piece tablebases
Ponder off
Balanced-12.ctg book
40 moves in 54 minutes repeating (adapted for the CCRL 40/40)
2 cycles (30 rounds)
Updated versions will not be introduced during the tournament.
The top ten engines are likely to be promoted into Division 5 for next season while the bottom six will likely drop into a Qualifier.
Athlon XP2800+
Fritz 10 GUI
128mb hash each
3-4-5 piece tablebases
Ponder off
Balanced-12.ctg book
40 moves in 54 minutes repeating (adapted for the CCRL 40/40)
2 cycles (30 rounds)
Updated versions will not be introduced during the tournament.
The top ten engines are likely to be promoted into Division 5 for next season while the bottom six will likely drop into a Qualifier.
Athlon XP2800+
Fritz 10 GUI
128mb hash each
3-4-5 piece tablebases
Ponder off
Balanced-12.ctg book
40 moves in 54 minutes repeating (adapted for the CCRL 40/40)
2 cycles (30 rounds)
Updated versions will not be introduced during the tournament.
The top ten engines are likely to be promoted into Division 5 for next season while the bottom six will likely drop into a Qualifier.
Athlon XP2800+
Fritz 10 GUI
128mb hash each
3-4-5 piece tablebases
Ponder off
Balanced-12.ctg book
40 moves in 54 minutes repeating (adapted for the CCRL 40/40)
2 cycles (30 rounds)
The top ten engines are likely to be promoted into Division 5 for next season while the bottom six will likely drop into a Qualifier.
Very disappointing performance by Joker. I have no idea why, maybe it is just bad luck. In Fonzy's tournaments, which have similar time controls, this version almost always ends above Buzz. In Scales of Justice, division 2, it kept on par with a set of much stronger engines.
Perhaps it hardly matters for Joker who the opponent is, so that it scores both 50% in a group of 2500-rated engines as wel as in a group of 2000-rated engines. This would explain the better performance at Fonzy's, as some very strong engines participate there.
Another possible explanation is that this version does not perform well on AMD CPUs at this time control (due to the small number of L1 cache ways on those machine, and an unluck cache collission between stacks and global data). I have seen that before on Pentium IV (which also has only 2 L1 cache ways), where at certain search depths the nps suddenly is cut in half.
Thanks for running the tournament Graham. Nice to see Garbochess doing decently . The C++ rewrite (currently written in C#) will hopefully be out soon, and should see a significant ELO boost.
gladius wrote:Thanks for running the tournament Graham. Nice to see Garbochess doing decently . The C++ rewrite (currently written in C#) will hopefully be out soon, and should see a significant ELO boost.