Author: Robert Hyatt
Date: 20:10:11 03/17/02
Go up one level in this thread
On March 17, 2002 at 02:25:25, Slater Wold wrote: A quick note... I played with this a bit. White is winning. period. I think that you _might_ simply be seeing hash table artifacts. IE crank it up and let it search a long time. Moves like b6 will fail low. The knight capture will fail low. _everything_ will fail low. What can happen in a game, with an SMP search, is that it can "see" more than it should because of old stuff in the hash table from the previous search. When you re-start, you don't see this old stuff and the scores change. I _often_ see cases in games where it gets a fail low at ply=N, while when I start it over, it won't fail low until ply=N+1 or N+2 (or more). That is _probably_ (but not positively) what is going on here.. >This position came up on an ICC game: > >[D]r1b2rk1/pp1n1ppp/4pB2/2R1P3/8/3B4/q4PPP/3Q1RK1 b - - 0 17 > >This is from my log: > >White(17): Bf6 > time used: 31.43 > time limit 14.74 (1:39) > nss depth time score variation (1) > 7 0.21 ++ 17. ... Re8!! > 7 0.35 -0.87 17. ... Re8 18. Qg4 g6 19. Bb5 Nxc5 > 20. Bxe8 Bd7 21. Bxd7 Nxd7 > 7 0.72 ++ 17. ... gxf6!! > 7 1.33 -3.42 17. ... Qa3 18. Qg4 g6 19. Rxc8 Rfxc8 > 20. Be4 Nxf6 21. exf6 Qb3 > (4) 7-> 1.39 -3.42 17. ... Qa3 18. Qg4 g6 19. Rxc8 Rfxc8 > 20. Be4 Nxf6 21. exf6 Qb3 > 8 1.66 -- 17. ... Qa3 > (3) 8 2.00 0.00 17. ... Qa3 18. Bxh7+ Kxh7 19. Qh5+ > Kg8 20. Bxg7 Kxg7 21. Qg5+ Kh7 22. > Qh5+ Kg7 > 8 2.64 ++ 17. ... Re8!! > 8 4.02 -0.93 17. ... Re8 18. Qh5 h6 19. Qg4 g6 20. > Rc7 Qd5 21. Qg3 Nxf6 22. exf6 > 8-> 5.61 -0.93 17. ... Re8 18. Qh5 h6 19. Qg4 g6 20. > Rc7 Qd5 21. Qg3 Nxf6 22. exf6 > 9 6.35 -- 17. ... Re8 > 9 16.54 5.92 17. ... Re8 18. Bxh7+ Kh8 19. Qh5 Qe2 > 20. Qxf7 gxf6 21. Qxe8+ Kxh7 22. Rxc8 > Nxe5 23. Qg8+ Kh6 24. Rxa8 > 9 16.88 ++ 17. ... Qa3!! > time=17.11 cpu=196% mat=-2 n=24143380 fh=94% nps=1411k > ext-> chk=1598422 cap=47309 pp=18507 1rep=176270 mate=14056 >sing=10 > predicted=16 nodes=24143380 evals=2438674 > endgame tablebase-> probes done=0 successful=0 > hashing-> trans/ref=28% pawn=99% used=65% > SMP-> split=482 stop=108 data=8/32 cpu=33.65 elap=17.11 > >I don't think there is a worse move than Qa3. I cringed when I saw this move >played. > >Oh well, that's why I am playing there, to catch these little things. So I go >back to my engine, and feed it the position, and let it think for 60 seconds. > >Here is the eval: > >White(1): setboard r1b2rk1/pp1n1ppp/4pB2/2R1P3/8/3B4/q4PPP/3Q1RK1 b - - 0 17 >Black(1): go > clearing hash tables > time surplus 0.00 time limit 1:00 (1:00) > nss depth time score variation (1) >starting thread 1 > (2) 7-> 0.25 -7.16 1. ... Nxc5 2. Qg4 g6 3. Bc4 Qc2 4. > Rd1 Qf5 > 8 0.39 -- 1. ... Nxc5 > 8 0.73 Mat06 1. ... Nxc5 2. Bxh7+ Kxh7 3. Qh5+ Kg8 > 4. Qg4 g6 5. Qh4 Qxf2+ 6. Rxf2 Nd7 > 7. Qh8# > 8 1.11 ++ 1. ... gxf6!! > 8 2.86 0.00 1. ... gxf6 2. Bxh7+ Kxh7 3. Qh5+ Kg8 > 4. Qg4+ Kh7 5. Qh5+ > 8 3.16 ++ 1. ... Re8!! > 8 3.98 -0.85 1. ... Re8 2. Bb5 Nxc5 3. Qg4 g6 4. > Bxe8 Nd7 5. Bxd7 Bxd7 > (2) 8-> 6.33 -0.85 1. ... Re8 2. Bb5 Nxc5 3. Qg4 g6 4. > Bxe8 Nd7 5. Bxd7 Bxd7 > 9 7.63 -- 1. ... Re8 > 9 8.97 4.60 1. ... Re8 2. Bxh7+ Kxh7 3. Qh5+ Kg8 > 4. Qg5 Nxf6 5. exf6 Qxf2+ 6. Kxf2 g6 > 7. Rc7 Rd8 8. Re7 a5 > 9 10.55 ++ 1. ... gxf6!! > 9 16.50 1.16 1. ... gxf6 2. Bxh7+ Kh8 3. Qh5 Qe2 > 4. Qh6 Qd2 5. Qxd2 Nxc5 6. Qh6 > 9 19.80 ++ 1. ... Nxf6!! > 9 21.38 -0.38 1. ... Nxf6 2. exf6 Qb2 3. Bxh7+ Kxh7 > 4. Qh5+ Kg8 5. fxg7 Qxg7 6. Rg5 f6 > 7. Rxg7+ Kxg7 8. Rd1 a5 9. g3 > (4) 9-> 24.44 -0.38 1. ... Nxf6 2. exf6 Qb2 3. Bxh7+ Kxh7 > 4. Qh5+ Kg8 5. fxg7 Qxg7 6. Rg5 f6 > 7. Rxg7+ Kxg7 8. Rd1 a5 9. g3 > (3) 10 28.78 -0.37 1. ... Nxf6 2. exf6 Qb2 3. Bxh7+ Kxh7 > 4. Qh5+ Kg8 5. fxg7 Qxg7 6. Rg5 f6 > 7. Rxg7+ Kxg7 8. Rd1 a5 9. f4 a4 10. > Kf2 > 10-> 36.83 -0.37 1. ... Nxf6 2. exf6 Qb2 3. Bxh7+ Kxh7 > 4. Qh5+ Kg8 5. fxg7 Qxg7 6. Rg5 f6 > 7. Rxg7+ Kxg7 8. Rd1 a5 9. f4 a4 10. > Kf2 > 11 51.30 -0.22 1. ... Nxf6 2. exf6 Qb2 3. Bxh7+ Kxh7 > 4. Qh5+ Kg8 5. fxg7 Qxg7 6. Rg5 f6 > 7. Rxg7+ Kxg7 8. Qg4+ Kf7 9. Qh5+ Ke7 > 10. Qc5+ Ke8 11. Rd1 Kf7 12. f4 > time=1:00 cpu=234% mat=-2 n=82822082 fh=90% nps=1372k > ext-> chk=3645381 cap=287364 pp=75218 1rep=496025 mate=38058 sing= >25 > predicted=0 nodes=82822082 evals=16376870 > endgame tablebase-> probes done=0 successful=0 > hashing-> trans/ref=33% pawn=28% used=96% > SMP-> split=1927 stop=605 data=7/32 cpu=2:21 elap=1:00 > > >Never even evaluates it. Strange!! > >Same version, settings, everything. I know it *probably* has to do with hash >and other in-game conditions. (And yes, positional learning _IS_ disabled.) >But it's *very* hard to diagnose a problem that cannot be reproduced. :( > >Anything but Nxf6 here is just wrong. Most programs will evaluate Qa3 (and >other losers) but fail low quickly on them, and move on to Nxf6.
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.