Author: Roy Brunjes
Date: 06:27:27 01/02/03
Go up one level in this thread
On January 02, 2003 at 08:53:35, scott farrell wrote: >This position arose in a game between chompster and a human kjf on ICC. > >kjf ran his clock down to about 6 secs in a 5 1 blitz game, my chompster had >4:30 on the clock, and lost !!! > >This is the position after Nxb2 ... blunder .... > >[D] r3qrk1/pp1b1ppp/8/4p1Q1/4N3/3B4/PnP3PP/2K2R2 w - - 0 19 > >The trick is to avoid taking the black queen, and instead Bxh7. > >Can some people let me know how nodes and what ply they solve this in. Any >inside into extensions, null move/pruning exclusions that helped would be >interesting. > >A slightly older crafty (18.14) needs about a guestimated 1.9 million nodes for >the final score, and about 600K nodes to see the fail-high, at depth 8. > >My chompster engine needs 1 million nodes to see the fail high, then 1.7million >nodes to see a score of over +3. > >my engine didnt see this tactic in the previous move, and doesnt see it until >some 6 million nodes, too many for my slow 100Knps program to see in a blitz >game. I am still trying to work out why it even wanted to play Nxb2 on the >previous move, I think it was trying to avoid a large number of checking moves, >quite clearly f5,h6,f6 are all better/safer/simpler. > >BTW does anyone know how to get a nice analysis out of CM8000 - all I can get is >a few english phrases which are nice, but no score/depth/node count etc. CM8000 >would only tell me about swapping queens for knights and bishops and such like, >and never did Bxh7, but I never got it to run proper analysis ... > >Thanx >Scott Here is the analysis from Junior 7 on this problem (64 MB hashtables, Pentium 4 1.6 Ghz): New game r3qrk1/pp1b1ppp/8/4p1Q1/4N3/3B4/PnP3PP/2K2R2 w - - 0 1 Analysis by Junior 7: 1.a4 Nxd3+ 2.cxd3 Bxa4 -+ (-4.88) Depth: 3 00:00:00 1.a3 Nxd3+ 2.cxd3 -+ (-4.55) Depth: 3 00:00:00 1.g4 Nxd3+ 2.cxd3 -+ (-4.49) Depth: 3 00:00:00 1.Nf6+ Kh8 2.Nxe8 Nxd3+ 3.cxd3 Raxe8 ³ (-0.29) Depth: 3 00:00:00 1.Nf6+ Kh8 2.Nxe8 Nxd3+ 3.Kb1 Raxe8 4.cxd3 h6 ³ (-0.29) Depth: 6 00:00:00 5kN 1.Nf6+ Kh8 2.Bxh7 gxf6 3.Qxf6+ Kxh7 4.Qh4+ Kg8 5.Qg5+ Kh8 6.Qf6+ Kh7 = (0.00) Depth: 9 00:00:00 196kN 1.Nf6+! ² (0.30) Depth: 12 00:00:01 687kN 1.Nf6+ Kh8 2.Bxh7 Qe6 3.Bg6 Nd3+ 4.Bxd3 Qxf6 +- (3.14) Depth: 12 00:00:06 4274kN 1.Nf6+ Kh8 2.Bxh7 Qe6 3.Bg6 Nd3+ 4.Bxd3 Qxf6 5.Rxf6 Kg8 6.Rd6 Bc6 7.Qxe5 Bxg2 8.Rd7 +- (3.30) Depth: 14 00:00:19 11036kN 1.Nf6+! +- (3.60) Depth: 15 00:00:30 17682kN 1.Nf6+! +- (3.90) Depth: 16 00:01:23 45733kN 1.Nf6+! +- (4.20) Depth: 17 00:03:00 94915kN 1.Nf6+! +- (4.50) Depth: 18 00:06:19 202282kN (Brunjes, MyTown 02.01.2003) And here is the output from CM9000: Time Depth Score Positions Moves 0:00 1/4 -0.96 10454 1.Nf6+ Kh8 2.Nxe8 Nxd3+ 3.cxd3 Raxe8 4.Qe3 Rc8+ 5.Kd2 0:00 1/5 0.00 58762 1.Nf6+ Kh8 2.Bxh7 gxf6 3.Qxf6+ Kxh7 4.Qh4+ Kg8 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ 0:01 1/6 2.86 222219 1.Nf6+ Kh8 2.Bxh7 Qe7 3.Bg6 Nd3+ 4.Bxd3 Qa3+ 5.Kd1 Qxd3+ 6.cxd3 gxf6 7.Qxf6+ Kg8 8.Qg5+ Kh8 9.Qxe5+ Kg8 0:04 1/7 2.74 502416 1.Nf6+ Kh8 2.Bxh7 Qe7 3.Bg6 Nd3+ 4.Bxd3 Qa3+ 5.Kd2 Qb4+ 6.c3 Qb2+ 7.Bc2 Qxc2+ 8.Kxc2 gxf6 9.Qxf6+ Kg8 10.Qg5+ Kh8 11.Qxe5+ Kg8 0:12 1/8 7.09 1633250 1.Nf6+ Kh8 2.Bxh7 Qe7 3.Qh5 Bg4 4.Nxg4 Qg5+ 5.Qxg5 f6 6.Qd2 Nc4 0:27 1/9 Mate11 4228250 1.Nf6+ Kh8 2.Bxh7 Qe7 3.Bg6 Nd3+ 4.Bxd3 Qa3+ 5.Kd1 Qxd3+ 6.cxd3 gxf6 7.Rxf6 Bg4+ 8.Qxg4 Kh7 9.Qh5+ Kg8 10.Qg5+ Kh8 11.Rh6# 0:38 1/10 Mate08 6141384 1.Nf6+ Kh8 2.Bxh7 Qe7 3.Qh4 Qxf6 4.Rxf6 Rfc8 5.Bg6+ Kg8 6.Qh7+ Kf8 7.Rxf7+ Ke8 8.Qg8# As you can see, CM9000 shows up Junior 7 here and while both show winning scores, mate is a more convincing score than 4.5 pawns! By the way, one way to get this output from CM9000 (not sure it is any different for CM8000) is to bring up the "Mentor Lines" window (go to the "Game Room" and under the menu "Mentor" select "Mentor Lines" and the thinking process will be displayed in a window. Roy
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.