Author: Uri Blass
Date: 03:25:50 08/19/03
Go up one level in this thread
On August 19, 2003 at 05:21:36, Uri Blass wrote: >On August 19, 2003 at 04:59:48, Uri Blass wrote: > >>On August 19, 2003 at 04:07:07, Tord Romstad wrote: >> >>>On August 18, 2003 at 19:16:51, Thorsten Czub wrote: >>> >>>>ok - i replayed the game, but on a slower machine... >>>> >>>>rebels evals in this game: >>> >>>Gothmog's evals interspersed (this time on a very fast Athlon MP 2.4 GHz): >>> >>>>move time score move >>>>21. 21" +0.83 Bd4 in ply 8. >>> >>>21. 4" +0.49 in ply 7 >> >>I suggest that you give it more time because it may fail high on Ba7 that is the >>wrong move because Rbc8 probably leads to a draw that is too deep for movei to >>see in a reasonable time >> >>Here is movei's analysis on Athlon 1000Mhz: >> >>depth=10 -0.05 f2d4 b4d5 h5f7 g8f7 f5f6 g7g6 g5d5 d7e6 d5a5 b8d8 f1e1 f8d6 g2g3 >>Nodes: 6653765 NPS: 184570 >>Time: 00:00:36.05 >>depth=10 +0.13 f2a7 b8c8 a7d4 b4d5 f5f6 d7e6 f6g7 f8e7 g5h6 c2c7 >>Nodes: 13410064 NPS: 188264 >>Time: 00:01:11.23 >>depth=11 +0.46 f2a7 f7f6 g5c1 c2c1 a1c1 b4d5 a7b8 e8b8 f1d1 d7c6 c1c4 e4e3 >>Nodes: 40042526 NPS: 189828 >>Time: 00:03:30.94 >>depth=11 +0.46 f2a7 f7f6 g5c1 c2c1 a1c1 b4d5 a7b8 e8b8 f1d1 d7c6 c1c4 e4e3 >>Nodes: 45778857 NPS: 191832 >>Time: 00:03:58.64 >>depth=12 +0.41 f2a7 f7f6 g5c1 b8c8 c1c2 b4c2 h5e8 c8e8 a1c1 d7a4 e2c3 a4b3 a7b6 >>f8d6 >>Nodes: 71178279 NPS: 191118 >>Time: 00:06:12.43 >> >> >>> >>>>22. 2'22" +1.36 (fail high!) Rac1 in ply 11. >>>> 3'54" +2.50 >>> >>>22. 12" +0.45 Rfc1 in ply 8. >>> >>>Rac1 is the really hard move to find. Even at ply 10, which takes about 5 >>>minutes >>>to complete, Gothmog keeps Rfc1 as its best move. The PV is 22. Rfc1 Qd3 23. >>>Ra3 h6 >>>24. Bxf7+ Kxf7 25. Qg6+ Kg8 26. Rxd3 exd3 27. Nc3 Nxc3 28. Rxc3 Re1 29. Bg1, >>>with a score of 0.40. >> >>Movei also does not find Rac1 in a reasonable time and prefers f6(I remember >>from previous analysis that it can find Rc3 easily but need minutes to find >>positive score at depth 11 or 12(I do not remember now) when the score was 0.00 >>before finding the 1.xx score. >> >>The problem was to find that Nxc3 is bad. >> >>Note that I did not implement complicated extension stuff and I prefer first to >>improve my pruning. >> >>Some analysis copied from winboard during writing this post: >> >>depth=10 +0.22 f5f6 g7g6 g5d5 g6h5 d5g5 g8h8 g5h5 c2b3 e2f4 f8d6 f4d5 d7c6 >>Nodes: 5730241 NPS: 184015 >>Time: 00:00:31.14 >>depth=11 +0.26 f5f6 g7g6 g5d5 g6h5 d5g5 g8h8 g5h5 c2b3 a4a5 f8d6 h1g1 e8g8 g2g3 >>Nodes: 15296240 NPS: 183540 >>Time: 00:01:23.34 >>depth=12 +0.33 f5f6 g7g6 g5d5 g6h5 d5g5 g8h8 g5h5 c2b3 e2f4 e8d8 a4a5 d7b5 f1d1 >>e4e3 >>Nodes: 33258364 NPS: 186425 >>Time: 00:02:58.40 >>depth=13 +0.27 f5f6 g7g6 g5d5 g6h5 d5h5 h7h6 f1d1 b8c8 d4e5 e8e5 h5e5 d7g4 e2d4 >>c2b2 >>Nodes: 112425992 NPS: 190226 >>Time: 00:09:51.01 > >I find that depth 13 is enough for movei to fail high on Rac1 > >depth=13 +0.57 a1c1 >Nodes: 351010457 NPS: 169284 >Time: 00:34:33.50 > >I guess that movei was slowed down by other application so the number of nps >went down > >Uri After solving the fail high it seems that more iterations are easy for movei and it shows the following analysis: depth=14 +0.88 a1c1 c2b3 c1c3 b3b4 f5f6 g7g6 g5d5 b8d8 c3c7 b4d6 c7d7 d8d7 d5d6 f8d6 h5g4 Nodes: 490863464 NPS: 168672 Time: 00:48:30.16 depth=15 +1.14 a1c1 c2b3 c1c3 b3b4 h5f7 g8f7 f5f6 d5f6 f1f6 f7g8 g5d5 g8h8 d5d7 b8d8 d7f7 b4b2 f6h6 b2a1 c3c1 a1a4 Nodes: 983598348 NPS: 174640 Time: 01:33:52.12 Uri
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.