Author: Thomas Mayer
Date: 09:33:56 01/22/02
Go up one level in this thread
Hi David, [D]8/r7/8/p2k4/R3p1K1/8/8/8 b - - 0 67 >Can your program find Kc5! (there maybe other winning moves, like Kc6), and can >it specifically avoid Ke5, which only draws? In other words, I don't know all >the moves that wins and all the moves that draws, but can your program find a >winning move without tablebases, and even understand what is going on? well, understandment of Quark might be low, but finds Kc5 anyway... Only the score does not really show that it understands that much... Analysis by Quark v1.68/A1400: 1...Ke5 2.Kg3 Tg7+ 3.Kf2 Kf4 4.Txa5 e3+ 5.Ke2 Tg2+ 6.Ke1 Ke4 7.Ta4+ Kd3 -+ (-3.06) Tiefe: 9 00:00:00 310kN 1...Ke5 2.Kg3 Tg7+ 3.Kf2 Kf4 4.Txa5 e3+ 5.Ke2 Tg2+ 6.Kd3 e2 7.Kd2 Kf3 8.Tf5+ Ke4 -+ (-2.96) Tiefe: 10 00:00:00 379kN 1...Ke5 2.Kg3 Ta6 3.Kf2 Kf4 4.Ke2 Th6 5.Txa5 Th2+ 6.Ke1 e3 7.Ta4+ Kf3 -+ (-3.02) Tiefe: 11 00:00:01 560kN 1...Ke5 2.Kg3 Ta6 3.Kf2 Kf4 4.Ke2 Th6 5.Txa5 Th2+ 6.Ke1 e3 7.Ta4+ Kf3 8.Ta7 -+ (-2.99) Tiefe: 12 00:00:01 1221kN 1...Ke5 2.Kg3 Tg7+ 3.Kf2 Kf4 4.Txa5 e3+ 5.Kf1 Kf3 6.Tf5+ Ke4 7.Tf8 Td7 8.Te8+ Kd3 9.Ke1 Tf7 -+ (-3.01) Tiefe: 13 00:00:03 2294kN 1...Ke5 2.Kg3 Tg7+ 3.Kf2 Kf4 4.Txa5 e3+ 5.Kf1 Td7 6.Ta8 Td1+ 7.Ke2 Td2+ 8.Ke1 Th2 -+ (-3.03) Tiefe: 14 00:00:05 3814kN 1...Ke5 2.Kg3 Tg7+ 3.Kf2 Kf4 4.Txa5 e3+ 5.Kf1 Td7 6.Ta2 Kf3 7.Tf2+ Ke4 8.Tb2 Th7 9.Kg2 Kd3 10.Kg3 Kd4 -+ (-3.04) Tiefe: 15 00:00:06 4798kN 1...Ke5 2.Kg3 Ta8 3.Kf2 Kf4 4.Ke2 Tb8 5.Txa5 Tb2+ 6.Kd1 e3 7.Ta7 Td2+ 8.Ke1 Tc2 9.Ta8 -+ (-3.06) Tiefe: 16 00:00:18 13975kN 1...Ke5 2.Kg3 Ta8 3.Kf2 Kf4 4.Ke2 Ta6 5.Kf2 Tb6 6.Txa5 e3+ 7.Ke2 Tb2+ 8.Ke1 -+ (-3.09) Tiefe: 17 00:00:23 17956kN 1...Ke5 2.Kg3 Ta8 3.Kf2 Kf4 4.Ke2 Ta6 5.Kf2 Tb6 6.Txa5 e3+ 7.Ke2 Tb2+ 8.Ke1 Tb1+ 9.Ke2 Tb2+ -+ (-3.09) Tiefe: 18 00:00:39 30125kN 1...Kc5 -+ (-3.10) Tiefe: 18 00:01:06 49968kN 1...Kc5 2.Kf4 Kb5 3.Txe4 a4 4.Te5+ Kc4 5.Te4+ Kd3 6.Te3+ Kc2 7.Te2+ Kc1 8.Kg4 a3 9.Ta2 Kb1 10.Tf2 Tg7+ 11.Kh3 a2 12.Tf1+ Kb2 13.Tf2+ Kb3 14.Tf3+ -+ (-3.14) Tiefe: 18 00:01:25 63718kN 1...Kc5 -+ (-3.44) Tiefe: 19 00:01:53 83957kN 1...Kc5 2.Kf4 Kb5 3.Txe4 a4 4.Te5+ Kc4 5.Te4+ Kd3 6.Te3+ Kc2 7.Te2+ Kc1 8.Te1+ Kd2 9.Ta1 a3 10.Ta2+ Kc3 11.Ke5 Kb3 12.Ta1 a2 13.Ke4 Te7+ 14.Kd4 Ka3 -+ (-3.95) Tiefe: 19 00:03:06 138793kN 1...Kc5 2.Kf4 Kb5 3.Txe4 a4 4.Te5+ Kc4 5.Te4+ Kd3 6.Te3+ Kc2 7.Te2+ Kc1 8.Te1+ Kd2 9.Ta1 a3 10.Ta2+ Kc3 11.Kf3 Kb3 12.Te2 Tc7 13.Te6 Tc3+ 14.Ke4 a2 -+ (-3.86) Tiefe: 20 00:04:55 228517kN In my opinion it is important that the program can find such things without Tablebases, I would really like it when Quark would find it as fast as Tiger or ChessMaster. Why ? There are many positions where you are not using Tablebases at all, e.g. in qsearch. The better your eval there is the better it will play, Tablebases are fine, but should not stop you from implementing knowledge of endgame, the more knowledge you have here, the lesser you must check Tablebases and that saves time... Ah, at ply 21 Quark gets up to over -4.xx (view of white) so it might have understood it now finally ?! Greets, Thomas
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.