Article 27741 of rec.games.chess.computer: Path: matra.meer.net!news.spies.com!news.sgi.com!newsfeed.nacamar.de!news.apfel.de!news-fra1.dfn.de!news-koe1.dfn.de!news.prima.ruhr.de!not-for-mail From: mclane@prima.ruhr.de (mclane) Newsgroups: rec.games.chess.computer Subject: The new YAZGAC - test-suite Date: Tue, 15 Apr 1997 09:36:31 GMT Organization: Prima e.V. Lines: 255 Message-ID: <5iviai$qma$2@steve.prima.ruhr.de> NNTP-Posting-Host: mclane.prima.ruhr.de X-Newsreader: Forte Free Agent 1.0.82 Xref: matra.meer.net rec.games.chess.computer:27741 Hello, the famous Computer Schach and Spiele edition 2/97 is on the market. Again mathematician and Dipl.Ing. YAZGAC has published an incredible test. It has 40 positions. I have hacked in these positions and will give you the epd's for CSTAL (because chris is so kind still using a non-epd-conform own format! GRRRRRRRRR!) and for normal-epd-programs: You have to run the positions and write down the time in seconds for the key-moves (e.g. in CSTal format the key moves are behind the epd in algebraic-notation. Some positions have more than one key-move, chris format is in the moment not capable handling this...). 10 Minutes are the maximum computing time. Above this, you get sero points for any position. Each position has its own weightings from 20-110. A good idea! But: Yazgac is doing a strange way of finding out about the solution-time. For my own purposes I use the usual way, writing down the time when the program first finds the move, and hold it to the end. Formula for points is: points = weigthing * (1 - seconds / 600 ) You can find the weigthings in my list below. If you have a p0 value, you can either find the elo out of this points by using 2 other before-rated program-measurements out of a chart, or, because not any of us has Computer-Schach & Spiele, and I have no time to hack in the charts... use formula 5 (which is only working from 2000-2600 ELO.) All in all I think some of the positions are to easy for state of the art programs. Maybe Mr. Yazgac should update some of his tested programs, or buy a faster pc than his used P90 , so he will get better formula's and more working test-suites. But maybe for him formulas and beeing a mathematician is more important than having working values ?! Yaz2.epd: 4rk2/2p2prp/pq2b2N/1p6/8/2PR1Q2/PP4PP/5R1K w - - 2r2rk1/pp4pp/2n1pq2/2bp4/8/2N3PB/PP2PP1P/R2Q1RK1 w - - 8/4Pn2/7p/3N1k2/4q3/5pK1/3Q4/8 w - - 5rk1/1pp3pp/p1p5/2b5/3n2bq/2NPB3/PPPQ2PP/R3N1K1 b - - 1rb2rk1/3nqppp/p1n1p3/1p1pP3/5P2/2NBQN2/PPP3PP/2KR3R w - - 3R4/5bkp/6p1/3P1p2/3q4/Q5PP/P1r1r1B1/5R1K b - - r1b2rk1/3n1ppp/p7/Qp1q2B1/P1pPp3/2P1P3/2B2PPP/R3R1K1 w - - r4rk1/1b2bppp/ppq1pn2/2ppB3/5P2/1P1BP1N1/P1PPQ1PP/R4RK1 w - - rn1qr1k1/p3bppp/bp3n2/2p2N2/2Q5/3BB1N1/PPP2PPP/2KR3R w - - 3r2k1/pq1P1rb1/1p3ppp/2p5/1nQ2B1P/5N2/P4PP1/1R1R2K1 w - - 1k1rr3/1p3p1p/p1p5/2P3p1/Q2b2q1/1R1BB1P1/PP3P1P/6K1 w - - 3r1r1k/1b4p1/pb5p/2ppPp1q/2p1nP2/2P1BN1N/PPQ3PP/3R1R1K b - - r2q1rk1/pb3ppp/1p2pnn1/6B1/2BP4/P1NQ4/1P3PPP/3R1RK1 w - - r3k3/1p3p2/p2p1b1r/3R3P/4q3/2P5/PP1QB3/R2K4 w q - r1r2qk1/pR3pp1/4p3/6P1/8/4PQ2/5PP1/5RK1 b - - 3r2k1/5pp1/7p/pq2r3/R2P4/6P1/1P1R3P/Q5K1 b - - r3nrk1/ppq2pbp/2n1b1p1/4pP2/2P5/4BNN1/PP4PP/RQ2KB1R b QK - 8/p2Q2p1/1k3qbp/1pn5/2p4P/2P2P2/5BPK/5B2 w - - r2rb1k1/pp1n1ppp/2p1pq2/4N3/3P4/4Q1P1/PPP2PBP/2KR3R w - - q4r2/1bn1ppkp/3p1np1/1ppP4/r3P3/2PBQNNP/1P3PP1/3RR1K1 w - - 4k1r1/1b1r4/1nppNp2/1p3Pp1/1P2P1P1/2N3KR/2P5/7R w - - r4k2/2p5/P1pp1n1p/2r2qp1/Q7/6BP/2P2PP1/R3R1K1 w - - 4r1k1/2rb1pp1/p1n1p2p/1p1pP3/3N1P2/1P1BK1P1/P1P2R1P/R7 w - - 3r2k1/1pq1r1p1/2pnp1Bp/p2p4/3P1P2/4PR1P/PPQ3P1/2R3K1 w - - 2kr3r/ppp1qppp/2p2n2/2b2b2/2P1pP2/1P2P3/PBQPB1PP/RN2K2R b QK - r4rk1/pp2bppp/3p1n2/4pP2/2q1P3/2N1B3/PPP3PP/R3QRK1 w - - 2rq1rk1/pp2bpp1/2pn1n1p/3p4/3P1B1P/2NBP3/PPQ2PP1/R3K2R w QK - 1n3rk1/r4ppp/pp1q4/3p4/Q2P4/5N2/PP3PPP/2R1R1K1 w - - 6r1/3bbk2/4p2p/1p1pPp2/2pP1Pr1/P4K2/1P4NP/1R2B1R1 b - - 2nq2rb/3pr2k/1pp1p2p/p3Pp1B/3P1P2/PPN5/5Q1P/R5RK w - - 7k/8/p4n2/2pP1r2/2P1p2P/8/P2N1p2/3K3R b - - 8/8/8/8/p2K4/3Q4/kp3P2/8 w - - 8/8/8/5R2/4K1k1/pp6/8/8 w - - 8/5p2/p3p1p1/1pk1P2p/5P1P/1P1K2P1/P7/8 b - - 8/8/8/1P2kp2/P2p2p1/6P1/3K4/8 b - - 8/1K6/P7/4p3/4Pk2/2pBb3/8/8 w - - 8/3k4/4R3/1p1r1p2/p4P2/P1K1P3/1P6/8 w - - 3K4/3P1k2/8/8/8/2r5/8/4R3 w - - 8/N5pp/8/8/3knP2/3p2P1/P6P/2K5 b - - 8/5r2/3k1P2/3p4/1p1K4/1P6/5R2/8 w - - Yaz2.pos: 40 positions : 40 Yazgac-Test. BT=2630 4rk2/2p2prp/pq2b2N/1p6/8/2PR1Q2/PP4PP/5R1K/w d3d7 2r2rk1/pp4pp/2n1pq2/2bp4/8/2N3PB/PP2PP1P/R2Q1RK1/w c3d5 8/4Pn2/7p/3N1k2/4q3/5pK1/3Q4/8/w e7e8Q 5rk1/1pp3pp/p1p5/2b5/3n2bq/2NPB3/PPPQ2PP/R3N1K1/b h4h6 1rb2rk1/3nqppp/p1n1p3/1p1pP3/5P2/2NBQN2/PPP3PP/2KR3R/w d3h7 3R4/5bkp/6p1/3P1p2/3q4/Q5PP/P1r1r1B1/5R1K/b f7e8 r1b2rk1/3n1ppp/p7/Qp1q2B1/P1pPp3/2P1P3/2B2PPP/R3R1K1/w a4b5 r4rk1/1b2bppp/ppq1pn2/2ppB3/5P2/1P1BP1N1/P1PPQ1PP/R4RK1/w g3h5 rn1qr1k1/p3bppp/bp3n2/2p2N2/2Q5/3BB1N1/PPP2PPP/2KR3R/w f5h6 3r2k1/pq1P1rb1/1p3ppp/2p5/1nQ2B1P/5N2/P4PP1/1R1R2K1/w b1b4 1k1rr3/1p3p1p/p1p5/2P3p1/Q2b2q1/1R1BB1P1/PP3P1P/6K1/w b3b7 3r1r1k/1b4p1/pb5p/2ppPp1q/2p1nP2/2P1BN1N/PPQ3PP/3R1R1K/b d5d4 r2q1rk1/pb3ppp/1p2pnn1/6B1/2BP4/P1NQ4/1P3PPP/3R1RK1/w f2f4 r3k3/1p3p2/p2p1b1r/3R3P/4q3/2P5/PP1QB3/R2K4/w d5d3 r1r2qk1/pR3pp1/4p3/6P1/8/4PQ2/5PP1/5RK1/b a7a5 3r2k1/5pp1/7p/pq2r3/R2P4/6P1/1P1R3P/Q5K1/b b5e8 r3nrk1/ppq2pbp/2n1b1p1/4pP2/2P5/4BNN1/PP4PP/RQ2KB1R/b e5e4 8/p2Q2p1/1k3qbp/1pn5/2p4P/2P2P2/5BPK/5B2/w g2g3 r2rb1k1/pp1n1ppp/2p1pq2/4N3/3P4/4Q1P1/PPP2PBP/2KR3R/w f2f4 q4r2/1bn1ppkp/3p1np1/1ppP4/r3P3/2PBQNNP/1P3PP1/3RR1K1/w e4e5 4k1r1/1b1r4/1nppNp2/1p3Pp1/1P2P1P1/2N3KR/2P5/7R/w e4e5 r4k2/2p5/P1pp1n1p/2r2qp1/Q7/6BP/2P2PP1/R3R1K1/w c2c4 4r1k1/2rb1pp1/p1n1p2p/1p1pP3/3N1P2/1P1BK1P1/P1P2R1P/R7/w d4c6 3r2k1/1pq1r1p1/2pnp1Bp/p2p4/3P1P2/4PR1P/PPQ3P1/2R3K1/w g2g4 2kr3r/ppp1qppp/2p2n2/2b2b2/2P1pP2/1P2P3/PBQPB1PP/RN2K2R/b f6g4 r4rk1/pp2bppp/3p1n2/4pP2/2q1P3/2N1B3/PPP3PP/R3QRK1/w e3g5 2rq1rk1/pp2bpp1/2pn1n1p/3p4/3P1B1P/2NBP3/PPQ2PP1/R3K2R/w g2g4 1n3rk1/r4ppp/pp1q4/3p4/Q2P4/5N2/PP3PPP/2R1R1K1/w a4c2 6r1/3bbk2/4p2p/1p1pPp2/2pP1Pr1/P4K2/1P4NP/1R2B1R1/b b5b4 2nq2rb/3pr2k/1pp1p2p/p3Pp1B/3P1P2/PPN5/5Q1P/R5RK/w c3e4 7k/8/p4n2/2pP1r2/2P1p2P/8/P2N1p2/3K3R/b e4e3 8/8/8/8/p2K4/3Q4/kp3P2/8/w d3c2 8/8/8/5R2/4K1k1/pp6/8/8/w f5f1 8/5p2/p3p1p1/1pk1P2p/5P1P/1P1K2P1/P7/8/b c5b4 8/8/8/1P2kp2/P2p2p1/6P1/3K4/8/b f5f4 8/1K6/P7/4p3/4Pk2/2pBb3/8/8/w d3b1 8/3k4/4R3/1p1r1p2/p4P2/P1K1P3/1P6/8/w e6e5 3K4/3P1k2/8/8/8/2r5/8/4R3/w e1f1 8/N5pp/8/8/3knP2/3p2P1/P6P/2K5/b e4d6 8/5r2/3k1P2/3p4/1p1K4/1P6/5R2/8/w f2a2 positions with >1 key-moves are: 4 h4g5, h4h6 24 g2g4, g1h1, g1h2 all other key-moves can be taken from the algebraic-list of yaz2.pos The magnificent formulas that came with a 3 pages explanation can be found on page 49-51. I wonder why a shallow magazine like CCS reports 3 pages about formulas, but I think this was done to impress the readers, to show how high-standarized the value of the CSS writers is (Hello Dieter, Hello Pitters!) ! I am impressed too. To find 3 pages formula about a non-topic formula definition any pupil knows from ZIRKELTRAINING in school-sports (weightings of LIEGESTUETZE compared to weigthings of ARMDRUECKEN), but only to find 1 1/2 pages about the Paderborn-tournament. What is more important ? CSS shows what is important here (for them) ! Anyway...., or - as my friend Jaap van den Herik always says: NEVERTHELESS.... :-) I have done this test on my Penitum120 machine with a Version 224 of Chess System Tal, because the Diogenes-programmer (impressed that his program was tested by somebody at all) told me (although his program is not strong, he knows enough about chess - so I trusted his advises, and we confirmed that the test looks interesting...:-)) that the test seem to represent good ELO-values. I have the SENDINFO-FILES that show in evidence the solution times of cstal with me, you can get them on demand. Moritz, can you make a post about this article in rgcc, together with the positions and maybe with some formulas, and FAQ or HOW TO DO instructions ?? I have not the time because of AEGON. Chris: I have many problems with the fact that CSTal is still not capable to find position 38 Rf1+. I thought we spoke about these EASY endgames often often often often. Maybe not often enough.... GRRRRRR! Lazy britains ! :-) Here are the results of Chess System Tal: (To differenciate this yazgac-test from his older predecessor, I call the new one no. 2. , a time in [brackets] means, it is over 10' and was too long... no points for that, but maybe on a faster machine... ) Yazgac2: 1-40 1"-10' 20-110 No. time points weigthings 1 1" 19,96 20 2 6'5" 7,83 20 3 21" 19,3 20 4 13" 19,56 20 5 0" 20 20 6 0" 30 30 7 6" 39,6 50 8 7' 15 60 9 5'44" 25,6 60 10 [11'36"] - 70 11 4" 69,53 70 12 - - 110 13 - - 20 14 26" 19,13 20 15 1'59" 24,05 30 16 4" 29,8 30 17 1'53" 40,58 50 18 [12'21"] - 50 19 - - 70 20 4" 69,53 70 21 4'35" 43,33 80 22 - - 90 23 0" 100 100 24 - - 100 25 14" 97,6 100 26 8" 108,53 110 27 - - 110 28 4" 109,26 110 29 - - 110 30 9'33" 4,95 110 31 14" 19,53 20 32 12" 19,6 20 33 7'14" 5,53 20 34 0" 20 20 35 1" 29,95 30 36 1'58" 24,1 30 37 - - 40 38 - - 60 39 9'20" 4,66 70 40 2'41" 58,53 80 So the P0 points of CSTal is 1095,04 ! (if I haven't done a mistake...) With formula 5 you should be (normally) able to calculate the elo: ELO = 1050 + 5,817 * p0 - 0,008847 * p0*p0 + 0,0000046 * p0*p0*p0 When I input this into the formula 5 , and use a very good calculator, I get 2851.144 ELO. Because Yazgac claims that his formulas work for ELO-ranges between 2000-2600 we see that we cannot USE this elo here. Ok, forget about the elo. The points are high enough. Now I am interested in your comments, ELO/points of Hiarcs6, Shredder, Mchess6 and Genius5 (Rebel8 and MCP5 , Fritz4 and Fritz3 and Wchess were tested). Thanks. I hope I have not made any (typing-/calculating-) mistake somewhere , but would be pleased if you tell me, if I have made a wrong position or wrong value somewhere. Critics recommanded.