Author: Robert Hyatt
Date: 07:27:54 01/29/04
Go up one level in this thread
On January 28, 2004 at 11:25:00, Mike S. wrote: >On January 27, 2004 at 14:49:12, Thomas Mayer wrote: > >>(...) And as far as I know they did test some positions in a one ply >>search... (Where the search itself does not play such a big role - mainly the >>eval decides about the move choosed in a one ply search...) - and the version of >>List they did test showed almost always the same score and move... > >I had not read that yet, which is new info for me, before posting a reply below >where I suggested just that (IOW. that this should have been done first). > >Now, I've done just this using the oldest versions of Crafty and List I had >quickly available. I think the theoretical, or suspected, chance of similarities >(or identical output) in such a case is bigger than when newer, further >developed engines are used for the comparison. These were cb-native Crafty 17.14 >(run in Fritz 8 GUI) and List 4.61 (WB., run in console mode using the depth and >test options). I took 10 positional test positions, because I thought the chance >of identical moves is generally better then - so that we can compare evals of >the same moves - than in "common" non-test positions. The ply depth was set to 1 >ply each. I give the EPD of the positions at the end of the posting. Some of >these pos. are somewhat doubtful in a common testing sense, but I think that >doesn't matter here. > >Result: > >From 10 positions, 6 identical moves were chosen and 4 moves were different. > >From the 6 same moves each, the evaluation ALWAYS was different. > >From the 6/10 same moves chosen by both, 3...5 are more or less "obvious" moves, >IOW. easy moves like defending the only attacked pawn or moving a piece out of >the the attack. Take a look at the moves and positions too see how that is >meant. I assume many engines would choose these moves in a one ply search >although I didn't check that. > >I think we can see this as exonerating evidence (=Entlastungsmaterial) in >addition to similar contributions by others in several places. > >Detailed results (evals from White's view): > >Pos.ID Crafty 17.14 List 4.61 >-------------------------------- >25 Bd7 0.17 Bd7 -0.07 >26 Rcd1 0.58 c5 0.73 >27 Rd7 -0.68 Rd7 -0.80 >28 Kf6 2.22 Kf6 2.49 >29 Bd7 -1.10 Bd7 -0.99 >30 Bxc5+ -0.34 Bxc5+ -0.86 >31 Bc6 0.59 f4 0.74 >32 Bg2 1.09 Re3 1.08 >33 Bb7 0.36 f6 1.00 >34 c4 1.41 c4 3.52 > > >Regards, >M.Scheidl > > >r1b2rk1/1p2qppp/1bn1p3/p3P3/8/PNQ3P1/1P1B1PBP/R4RK1 b - - bm a4; id "Mike's Test >2.2, Nr. 25"; >r5k1/p2r1bpp/2p2p2/8/n1P5/P5B1/5PPP/2R1RBK1 w - - bm c5; id "Mike's Test 2.2, >Nr. 26"; >3r1bk1/p4ppp/Qp2p3/8/1P1B4/Pq2P1P1/2r2P1P/R3R1K1 b - - bm e5; id "Mike's Test >2.2, Nr. 27"; >7k/4K1n1/8/6P1/8/3B4/8/8 w - - bm Bg6; id "Mike's Test 2.2, Nr. 28"; >r3nrk1/ppq2pbp/2n1b1p1/4pP2/2P5/4BNN1/PP4PP/RQ2KB1R b KQ - bm e4; id "Mike's >Test 2.2, Nr. 29"; >8/p2Q2p1/1k3qbp/1pn5/2p4P/2P2P2/5BPK/5B2 w - - bm g3; id "Mike's Test 2.2, Nr. >30"; >2r2rk1/1p1bq3/p3p2p/3pPpp1/1P1Q4/P7/2P2PPP/2R1RBK1 b - - bm Bb5; id "Mike's Test >2.2, Nr. 31"; >r1n1kb1r/1b3pp1/1qn1p3/p2pP1Bp/Qp1P1N1P/5NPB/PP3P2/2R1R1K1 w kq - bm Nxe6; id >"Mike's Test 2.2, Nr. 32"; >1r1r2k1/2b1qp1p/b1p3p1/p1p1p3/2P1P3/1PN1BP2/P1Q3PP/R2R2K1 b - - bm Rd4; id >"Mike's Test 2.2, Nr. 33"; >r4k2/2p5/P1pp1n1p/2r2qp1/Q7/6BP/2P2PP1/R3R1K1 w - - bm c4; id "Mike's Test 2.2, >Nr. 34"; I thought this was put to bed months ago by Dann? He (I thought) said he had seen the source and it was not crafty. That was good enough for me at the time. There are definitely a few crafty clones available on the net, which claim to be new and original programs, but I've seen nothing pointing to list that held up under any sort of scrutiny...
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.