Author: Stuart Cracraft
Date: 20:15:03 07/11/04
Go up one level in this thread
On July 11, 2004 at 20:52:26, Dieter Buerssner wrote: >On July 11, 2004 at 20:36:52, Stuart Cracraft wrote: > >>You'd probably simply do much better to implement square of the pawn >>with a couple hundred point strength gain. It would clarify a huge >>number of the nodes with Fine #70. If enemy king is outside of >>the square of your passed pawn and does not have the move, etc. > >Stuart, this sounds interesting, but perhaps testing just the other way around >will be better (for debugging). Fine 70 can be solved with material only eval. >After 26 plies, you can see, that you can force the winning of a pawn. So, I'd >suggest, to debug this problem, disable all your eval, just return material >score. Also, one might want to disable repetition detection (repetition does not >mix well with transposition tables, and to make the debugging problem as easy as >possible, disabling repetion detection might help). > >About the nuclear simulations of the other post. You might have misunderstood my >point slightly. There are very good pseudo random number generators available in >source code. Some have no failed any tests for randomness. That particilar >lagged fibonacci PRNG did fail some simple tests. Simulating some easy >statistics with a small and simple (but sufficient) program using that PRNG (a >typical "Monte Carlo simulation" can give totally wrong results. While the same >program will work perfectly with another PRNG. If in doubt, an one wants to do >some simulation, using Mersenne Twister will probably be a good choice. If one >wants to be really careful, combine Mersenne Twister with one or two other well >known 32-bit PRNGs, for example KISS of Marsaglia, and some combined linear >congruential generator. myrand32var = >mersennetwister()^kissprng()^sophisticated_lcg(); > >If you find any problem where such a PRNG fails (even when uncombined Mersenne >twister alone fails) in a correctly implemented simulation, your fame in the >area of PRNGs will be guaranteed :-) > >Of coruse, this is very off topic here. No problem -- I wondered about PRNG for a time. Regarding 26 ply in Fine 70, my problem is simply that my code with my hardware can't get there in a reasonable period of time. 4 1/2 seconds for 21 ply won't give an answer for 26 ply for quite a long time in my case. (22 ply = 55 seconds), 23 = 100 seconds, 24 = 123 seconds, 25 ply = 297 seconds, 26 ply = 780 seconds. And it sees the two pawns there as you predicted. But later at the 30th ply it sees only one pawn advantage. So my code must be non-optimal or the hardware is too slow or a combination. Here is everything from ply 13 on: 13. a1b2 0.09 1017 16619 a1b2 a7b8 b2c1 b8c8 c1d2 c8c7 d2e1 c7b8 e1f1 b8c8 f 1g1 c8b8 g1h1 14. a1b2 0.12 1017 21965 a1b2 a7b8 b2a1 b8c8 a1a2 c8b8 a2b3 b8a7 b3a2 a7b8 15. a1b2 0.14 1017 26615 a1b2 a7b8 b2a1 b8c8 a1a2 c8b8 a2b1 b8c8 b1a2 16. a1b2 0.37 1017 72771 a1b2 a7b8 b2a1 b8c8 a1b1 c8d8 b1b2 d8e8 b2c3 e8e7 c 3d3 e7f6 d3c3 f6g6 c3c4 g6h5 c4b5 h5g4 17. a1b2 0.46 1016 97150 a1b2 a7b8 b2a1 b8c8 a1a2 c8b8 a2b3 b8b7 b3c2 b7c8 c 2c1 c8b8 c1d1 b8c8 d1e1 c8d8 e1f1 d8e8 f1g1 e8f8 g1h1 f8g8 h1g1 g8h8 g1h1 h8g8 18. a1b2 0.88 1017 192288 a1b2 a7b8 b2c2 b8c8 c2d3 c8c7 d3c3 c7c8 c3c4 c8b8 c 4b5 b8c8 b5a5 c8b8 a5b4 b8c8 b4c3 c8b8 c3b2 b8c8 b2a1 c8b8 19. a1b2 1.67 1017 354578 a1b2 a7a8 b2a1 a8b8 a1b1 b8c8 b1a2 c8b8 a2b1 20. a1b2 2.76 1017 585884 a1b2 a7a8 b2c2 a8b8 c2b3 b8c7 b3c4 c7b6 c4b5 21. a1b2 4.51 1017 979913 a1b2 a7a8 b2c2 a8b8 c2c3 b8b7 c3c4 b7b6 c4b5 22. a1b2 55.36 1012 11751898 a1b2 a7a8 b2c2 a8b8 c2c3 b8b7 c3c4 b7b6 c4b5 23. a1b1 100.56 1024 18967187 a1b1 a7b7 b1c1 b7c7 c1d1 c7d8 d1c2 d8e8 c2d2 e8e7 d2c3 e7f7 c3c4 f7g6 c4b5 g6h5 b5a5 h5g4 a5b6 g4f4 b6b7 f4e3 b7c7 e3d4 c7d6 24. a1b1 123.31 1024 24268004 a1b1 a7b7 b1c1 b7c7 c1d1 c7d8 d1c2 d8e8 c2c3 e8f7 c3c4 f7g6 c4b5 g6h5 b5a5 h5g4 a5b6 g4f4 b6c7 f4e3 c7d6 e3d4 d6e6 d4e4 25. a1b1 297.82 1012 59933186 a1b1 a7b7 b1c1 b7c7 c1d1 c7d8 d1e1 d8d7 e1f2 d7e8 f2e2 e8d8 e2d3 d8c7 d3c2 c7b8 c2c3 b8b7 c3c4 b7b6 c4d3 b6c7 26. a1b1 780.92 2006 168201254 a1b1 a7b8 b1c2 b8b7 c2c3 b7c8 c3c4 c8d8 c4b5 d8e8 b5a5 e8f8 a5b6 f8g8 b6c7 g8h8 c7d6 h8g7 d6e7 g7h6 e7d8 h6h5 d8c8 h5g4 d5d6 g4f4 d6d7 27. a1b1 897.06 2006 186865396 a1b1 a7b8 b1c2 b8b7 c2c1 b7a7 c1d2 a7b8 d2e1 b8b7 e1f1 b7c7 f1g2 c7d8 g2g3 d8e8 g3h4 e8f7 h4g5 f7g8 g5f5 g8f7 f5e4 f7g6 e4e3 g6f6 e3e2 f6g7 e2f1 g7h8 f1g1 h8g8 g1g2 g8h8 g2h1 h8g8 h1g2 28. a1b1 899.09 2006 187182423 a1b1 a7b8 b1c2 b8c8 c2b2 c8b8 b2b1 b8c8 b1b2 29. a1b1 913.61 2006 189927255 a1b1 a7b8 b1c2 b8a7 c2d1 a7b8 d1e1 b8c8 e1d2 c8b8 d2e1 30. a1b1 925.41 1012 192097753 a1b1 a7b7 b1c2 b7b8 c2b3 b8a7 b3a2 a7b8 a2b2 b8c8 b2a1 c8b8 a1b1 b8b7 : Thanks, Stuart
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.