Author: Peter Kappler
Date: 23:03:04 03/10/99
Go up one level in this thread
On March 10, 1999 at 23:08:47, blass uri wrote: > >On March 10, 1999 at 22:59:37, Robert Hyatt wrote: > >>On March 10, 1999 at 22:49:00, Robert Hyatt wrote: >> >>>On March 10, 1999 at 21:54:51, blass uri wrote: >>> >>>> >>>>On March 10, 1999 at 21:16:33, Robert Hyatt wrote: >>>> >>>>>On March 10, 1999 at 02:33:14, Peter Kappler wrote: >>>>> >>>>>> >>>>>>8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - bm Bh3; id "Topalov-Shirov Linares 98"; >>>>>> >>>>>> >>>>>>Does everybody remember this game? It's from last year's Linares tournament, >>>>>>where Shirov played a shocking bishop sacrifice in the endgame and won >>>>>>brilliantly. I believe that post-mortem analysis proved it was the only way to >>>>>>win. (Please correct me if this is wrong) >>>>>> >>>>>>I watched this game live on ICC with dozens of others, and we were all quite >>>>>>impressed with Shirov's powers of calculation. >>>>>> >>>>>>This *should* be an extremely difficult problem for computers - but I'm sure >>>>>>somebody will tell me that Hiarcs or some other commercial program can solve it >>>>>>in 10 seconds. >>>>>> >>>>>>Just curious... >>>>>> >>>>>>--Peter >>>>> >>>>> >>>>>I ran this two ways.. First from the original position to see what crafty would >>>>>play with no 'urging' by me: >>>>> >>>>> 12-> 6.37 2.61 1. ... a3 2. Kf2 Bc2 3. g4 Bd1 4. h5 >>>>> gxh5 5. gxh5 Bxh5 6. Ke3 a2 7. Kd4 >>>>> Bd1 8. Ba1 f5 >>>>> 13 11.78 2.62 1. ... a3 2. Kf2 Bc2 3. g4 Bd1 4. h5 >>>>> gxh5 5. gxh5 Bxh5 6. Ke3 a2 7. Bb2 >>>>> Bg6 8. Kd4 Be4 9. Ba1 >>>>> 13-> 13.87 2.62 1. ... a3 2. Kf2 Bc2 3. g4 Bd1 4. h5 >>>>> gxh5 5. gxh5 Bxh5 6. Ke3 a2 7. Bb2 >>>>> Bg6 8. Kd4 Be4 9. Ba1 >>>>> 14 34.53 2.46 1. ... a3 2. Kf2 Bc2 3. Ba1 Kf5 4. >>>>> Ke3 Be4 5. g3 a2 6. Kf2 Bc2 >>>>> 14-> 45.46 2.46 1. ... a3 2. Kf2 Bc2 3. Ba1 Kf5 4. >>>>> Ke3 Be4 5. g3 a2 6. Kf2 Bc2 >>>>> 15 1:08 2.46 1. ... a3 2. Kf2 Bc2 3. Ba1 Kf5 4. >>>>> Ke3 Be4 5. g3 a2 6. Kf2 Bc2 >>>>> 15-> 1:24 2.46 1. ... a3 2. Kf2 Bc2 3. Ba1 Kf5 4. >>>>> Ke3 Be4 5. g3 a2 6. Kf2 Bc2 >>>>> 16 2:55 2.37 1. ... a3 2. Kf2 Bc2 3. Ke3 a2 4. g3 >>>>> Kf5 5. Kd4 Be4 6. Kc5 Kg4 7. Bxf6 Kxg3 >>>>> 8. Kd6 Kg4 >>>>> 16-> 3:59 2.37 1. ... a3 2. Kf2 Bc2 3. Ke3 a2 4. g3 >>>>> Kf5 5. Kd4 Be4 6. Kc5 Kg4 7. Bxf6 Kxg3 >>>>> 8. Kd6 Kg4 >>>>> 17 8:25 2.42 1. ... a3 2. Kf2 a2 3. Ke2 Be4 4. Kf2 >>>>> Kf5 5. g3 Bb1 6. Ke3 Bc2 7. Kd4 Be4 >>>>> 8. Kc5 Kg4 9. Bxf6 >>>>> >>>>> >>>>>so it likes a3 thru any depth it might get in a real tournament setting. >>>>> >>>>> >>>>>I then played Bh3 and let it search for white: >>>>> >>>>> 16-> 49.18 0.56 2. gxh3 Kf5 3. Kf2 Kf4 4. Bxf6 a3 5. >>>>> Bc3 Ke4 6. Ba1 Kf4 7. Bh8 >>>>> 17 1:14 0.68 2. gxh3 Kf5 3. Kf2 Kf4 4. Bxf6 a3 5. >>>>> Bc3 Ke4 6. Ke2 Kf4 7. Kd3 Kg3 8. Bf6 >>>>> a2 9. Kc2 Kf4 10. Kb3 Kf5 >>>>> 17-> 1:22 0.68 2. gxh3 Kf5 3. Kf2 Kf4 4. Bxf6 a3 5. >>>>> Bc3 Ke4 6. Ke2 Kf4 7. Kd3 Kg3 8. Bf6 >>>>> a2 9. Kc2 Kf4 10. Kb3 Kf5 >>>>> 18 1:56 -- 2. gxh3 >>>>> 18 2:39 0.00 2. gxh3 Kf5 3. Kf2 Ke4 4. Bxf6 a3 5. >>>>> h5 gxh5 6. Ba1 d4 7. Ke2 h4 8. Kd2 >>>>> Kf4 9. Bxd4 Kg3 10. Bc3 Kxh3 >>>>> 18-> 3:26 0.00 2. gxh3 Kf5 3. Kf2 Ke4 4. Bxf6 a3 5. >>>>> h5 gxh5 6. Ba1 d4 7. Ke2 h4 8. Kd2 >>>>> Kf4 9. Bxd4 Kg3 10. Bc3 Kxh3 >>>>> 19 4:37 0.00 2. gxh3 Kf5 3. Kf2 Ke4 4. Bxf6 a3 5. >>>>> h5 gxh5 6. Ba1 d4 7. Ke2 h4 8. Kd2 >>>>> Kf4 9. Bxd4 Kg3 10. Bc3 Kxh3 >>>>> 19-> 5:24 0.00 2. gxh3 Kf5 3. Kf2 Ke4 4. Bxf6 a3 5. >>>>> h5 gxh5 6. Ba1 d4 7. Ke2 h4 8. Kd2 >>>>> Kf4 9. Bxd4 Kg3 10. Bc3 Kxh3 >>>>> >>>>>which is right interesting, because notice this runs into the 5 piece >>>>>tablebases and says _draw_. >>>>> >>>>>question is, now, does Bh3 really make sense? This seems to say "no". >>>>> >>>>>interesting... >>>>Did you let crafty to play against itself. >>>>after 2.gxh3 Kf5 3.Kf2 Ke4 4.Bxf6 >>>>crafty16.5 as an engine for Junior5(no tablebases) prefers 4...d4 with >>>>evaluation -6.01 at depth 18 and -6.02 at depth 19 >>>> >>>>I believe that crafty needs more time to see the win after Bh3. >>>> >>>>I do not remember the game but I remember that I analyzed it with a computer and >>>>did not see a draw for white. >>>> >>>>Uri >>> >>> >>>I will check deeper.. but note that after Bh3, it reaches the endgame databases >>>and with white on move, it claims white can force a dead draw. I don't see any >>>reason it would change to let black do more, after forcing a path to a tablebase >>>draw... >>> >>>And note that crafty is using _all_ the 5 piece files we have, which covers >>>a lot of territory. But I'll let it run a ways longer forcing the 'drawing' >>>white move first to see if black can 'improve'. >> >> >>Still says 'draw'... but there is a minor problem here, in that crafty believes >>black is losing and so would choose a draw given the chance. And there is no >>way to solve this. >> >>What I might propose is that we simply play this out... I'll run crafty on the >>white side, you play the black side. And let's see if it is winnable against >>the database 'monster'. I'd suggest we start after Bh3 gxh3 (assuming gxh3 is >>the best move for now, since Crafty says that leads to a tablebase draw). >> >>It alternates between Kf5 and f5 as black's next move, but I'll leave that to >>you to choose if you want to try this. Seems like an interesting thing to do >>of course. I'd love to be able to write that his Bh3 was no more than a draw >>after his occasional comments about 'no computer will ever solve this position' >>only to have one find the right move in minutes. :) >> >>I have gone thru 21 plies with black's second move and crafty still says Kf5, >>score=0.00 >> >>Bob > >After 1...Bh3 2.gxh3 I play 2...Kf5 and if 3.Kf2 I play Ke4 and if 4.Bxf6 I play >4...d4 > >Uri That is exactly how the game continued... 47. ... Bh3!! 48. gxh3 Kf5 49. Kf2 Ke4 50. Bxf6 d4 51. Be7 Kd3 52. Bc5 Kc4 53. Be7 Kb3 and White resigned. According to Seirawan's notes in Inside Chess, Black will play 54...Kc2 and queen one of the pawns. So, can Crafty improve upon Topalov's play? --Peter
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.