Author: Uri Blass
Date: 04:26:07 09/21/03
Go up one level in this thread
On September 21, 2003 at 01:00:37, Paul Byrne wrote: >I read the thread on the pawn game last night... >http://www.talkchess.com/forums/1/message.html?316441 > >...and added it as a variant to Guildenstern (which is pretty basic >as chess engines go, but is set up to handle rules variations like >this -- first version took about 10 minutes). The first version >called it a win when a pawn hit the 7th rank; changing this to >getting a passed pawn more advanced than the opponent's most advanced >pawn (passed or not) caused a speedup of about 100x in solving the >couple of test positions I was using. > >Anyways, to give you an idea how solvable this game is or isn't, >I left it running on this position last night: > >[D]8/1pppppp1/8/p6p/P6P/8/1PPPPPP1/8 w - - 0 3 > >with this result: > >mv=g4 e=12 d=1 n=14 t=0.000 >mv=f4 e=0 d=2 n=104 t=0.000 >mv=f4 e=12 d=3 n=371 t=0.002 >mv=f4 e=0 d=4 n=870 t=0.002 >mv=f4 e=12 d=5 n=2859 t=0.006 >mv=f4 e=-7 d=6 n=7296 t=0.013 >mv=f4 e=12 d=7 n=17061 t=0.028 >mv=f3 e=-4 d=8 n=56297 t=0.128 >mv=f4 e=19 d=9 n=152838 t=0.351 >mv=f4 e=-7 d=10 n=253530 t=0.586 >mv=e4 e=19 d=11 n=718726 t=1.473 >mv=e4 e=-2 d=12 n=1169031 t=2.114 >mv=e4 e=19 d=13 n=1875217 t=3.137 >mv=e4 e=-2 d=14 n=3304028 t=5.235 >mv=e4 e=16 d=15 n=6028707 t=9.264 >mv=e4 e=0 d=16 n=14889769 t=22.461 >mv=e4 e=28 d=17 n=116308613 t=168.791 >mv=e4 e=0 d=18 n=277650527 t=400.234 >mv=e4 e=27 d=19 n=493734227 t=706.427 >mv=e4 e=0 d=20 n=830582334 t=1181.929 >mv=e4 e=14 d=21 n=1572079524 t=2209.587 >mv=e4 e=0 d=22 n=-989389009 t=4577.107 >mv=e4 e=32 d=23 n=1675198974 t=8170.032 >mv=e4 e=273 d=24 n=1318190910 t=13610.508 >mv=e4 e=9969 d=25 n=-1677339015 t=26967.618 > >So about 7.5 hours to solve this. > >I'm sure a program set up especially to solve this would be noticably >faster (no need for code for castling, promotion, non-pawn moves, etc, etc) >and running on a faster machine than my 1.2 GHz athlon would gain a ply or >two. > >Adding a better eval function would likely aid move ordering some, accounting >for the symmetry of the initial position would help, etc. There are plenty >of ways to improve on my basic program. > >I have to suspect the game is solvable -- perhaps not quickly, but not in a >totally unreasonable about of time. > >-paul I believe that extensions also make help you. single reply extension when there is only one move that is not losing has chances to save you some plies. I think that calculating this condition is not hard and you only need to see that every move except one allow the opponent a capture that leads to unstoppable passed pawn that is more advanced than the opponent pawns. 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.