1994 ACM International Computer Chess Championship Hal Bogner Rd 3: The two leaders (both 2-0) played, and Star-Socrates beat M-Chess Pro Deep Thought II beat WChess Zarkov beat Now Cray Blitz beat Innovation II Evaluator beat Spector As in human tournamnets, sometimes you have to play the fellow you are rooming with; Zarkov and Now were doing so, as were Spector and Evaluator leader: Star Socrates (3), DT, MChess, Zarkov (2)... Rd 4: Deep Thought II beat Star-Socrates in a brilliant game to grab a share of the lead, Zarkov and M-Chess Pro drew a long R&P ending WChess beat Cray Blitz Innovation II beat Spector in a wacky game with lots of hanging pieces Now beat Evaluator scores: 3-1 DT, Star-Soc 2.5 MChess Pro, Zarkov 2 Innovation II, Now, WChess 1.5 Cray Blitz 1 Evaluator .5 Spector Rd 5: Deep Thought II beat MChess Pro handily to guarantee at least a share of first Zarkov beat Star-Socrates in another long R&P ending, leaving DT II clear first, and taking clear second, Now beat Innovation II Cray Blitz clobbered Spector and Evaluator recovered from an apparent typo in its opening book (f3 in a double-KP opening!?) to upset WChess Final scores: 4-1 Deep Thought II 3.5 Zarkov 3 Star-Socrates, Now 2.5 Cray Blitz, MChess Pro 2 Wchess, Evaluator, Innovation II .5 Evaluator DT II was 4-0 in played games, though it was probably losing to Zarkov in the middlegame in Rd 1; it's "loss" was a forfeit when the IBM facility lost all power Saturday night prior to the 7 PM start time. The scheduled game against MChess stayed on hold until 10:45 PM, when Hsu finally gave up hopes of recovery. MChess Pro gave DT II a second chance in the last round.... Zarkov beat Star-Soc's monster machine, Cray's super-computer, and gave DT II a run for its money; all on a machine about 1.5x faster than the Pentiums. For me, this was very satisfying, having worked some with John Stanback since 1990. John does this in his spare time (he's an IC engineer at HP). He got some help from IM Marc Leski over the last six months, and it looks like it was worthwhile! Innovation scored its first points after going 0-5 last year (the first time it played in the ACM tournament) Evaluator and Spector played in this event for the first time. The games will be forthcoming soon! -hal bogner (usually unbiased...) In article jbum@netcom.com (Jim Bumgardner) writes: >Here are the specs on the machines which completed in the tournament. >Hopefully Hal or Steve will post the results for rounds 3-5 soon. > >Note that the nodes per second vary greatly from position to position >- they are provided to give you a sense of the order of magnitude. > > >Cray Blitz: Fortran + Cray assembly, 4 proc. C94 (1000 mip). 150k for >executable code + space for hash tables. Uses dynamic tree splitting, >a variation of PVS which minimizes idle time and sync wait times by >using a sort divide and conquer algorithm. Book 300K+. 750k nodes >per second. > >Deep Thought II: C + microassembly, IBM RS/6000 580, Memory space: >100 mb; 100 mips Runs on 12-16 processors. Endgame DB from Thompson's >CD-ROMS. 4000k nodes per second. Note: this is not the new "Deep >Blue" hardware, as some hoped to see, but the old hardware running >new software. > >Evaluator: C, IBM PC 486 DX4 100 Mhz, 8mb, 131k for executable code. >Book 15K positions. 7K nodes per second. > >Innovation II: C, Mac Power PC (some games were played in emulation mode due >to crashes, others native), 20mb Executable code: 435k, Book: 6500 entries, >Endgame DB: KPK, KQK, KRK, 4k nodes per second. > >M-Chess Pro: C + Intel asm, Pentium 90 mhz, 11mb, 200kb for executable >code, Book: 300k moves. 9k nodes per second. > >Now: Pascal, PC Clone, Pentium 90 mhz, Book: 7200 positions, hash table 256k >positions. KPK perfect endgame knowledge (rules, rather than table look up). > >Spector: C, PC Clone 486 66Mhz with 256kb level two cache, 11 mips, executable >code 200k, 3 meg for data, book 200k positions, 3k nodes per second. > >Star-Socrates: Threaded-C, Connection Machine CM-5, 512 processors, 16 gb, >Uses a parallelization of scout (pvs) search called Jamboree search. >Book 13k positions, KPK endgame database. 1000k nodes per second. > >WChess: C, 90 Mhz Pentium, 16 mb, 200kb for executable code, Book: 105k moves. >37k nodes per second. > >Zarkov: C, HP735, 87 mips, 175k executable code, book 17k positions; KPK >end game database. 256k pos hash table, 15k nodes per second. > > >-- >-- >-- Jim Bumgardner | jbum@netcom.com >-- Time Warner Interactive >-- GSC/MUd--p(-)c+++!lu+e-m*s/+n++(---)h--f!gw+t+ry++