Author: Robert Hyatt
Date: 06:45:43 02/04/04
Go up one level in this thread
On February 04, 2004 at 08:58:51, Jarkko Pesonen wrote: >Re: How many programs can select the 1......h6! drawish move ? >http://chessprogramming.org/cccsearch/ccc.php?art_id=295554 > >If there ever was a hole in crafty's code, was it fixed as you intended to do? > >Jarkko Yes. IE you really have two choices. If you notice that your program does not play king and pawn endings well, or enters a king and pawn ending that is lost but it doesn't recognize it, you can do one of two things. (1) hope it doesn't happen in an important game or (2) fix it. I don't worry about reaching king and pawn endings vs GM players. I rarely get tricked into entering a lost position (it can happen in positions that look winnable but are not due to some subtle tempo issue or whatever). I don't worry about trading or not trading, etc. I worry more about king safety as that is probably the most complex problem to handle, since it is easy to "self-immolate" when you start an attack that really hurts you more than your opponent. The hole in the code from the message below was looked at, I don't recall whether it was fixed or put off until another day however.. > > >Some programs would select to move the king or 1....h5? which allow white to >penetrate, but 1....h6! would be the correct move by just sitting tight, and >there is no danger of Whites's king penetrating. > >[D]8/1p2k2p/4p3/1PP2pp1/3P4/7P/5KP1/8 b - - 0 1 > > >Posted by : Robert Hyatt on May 07, 2003 at 10:46:49 > >On May 06, 2003 at 11:14:19, José Carlos wrote: > >>On May 06, 2003 at 08:55:07, Robert Hyatt wrote: >> >>>On May 06, 2003 at 07:05:02, Jarkko Pesonen wrote: >>> >>>>Crafty 19.3 ends up in this position >>>> white black >>>> 1 d5 e5 >>>> 2 c6+ bxc6 >>>> 3 bxc6+ Kc8 >>>> 4 d6 e4 >>>> 5 Kd4 h5 >>>> 6 g4 hxg4 >>>> 7 hxg4 f4 >>>> 8 Kxe4 ... >>>> >>>>[D]2k5/8/2PP4/6p1/4KpP1/8/8/8 b >>>> >>>> 21-> 1:40 6.47 1. ... Kd8 2. Kd4 Kc8 3. Kd3 Kd8 4. >>>> Ke4 Kc8 5. Kf3 Kd8 6. Kf2 Kc8 7. Ke2 >>>> Kd8 8. Kd3 Kc8 9. Kc3 Kd8 10. Kd2 Kc8 >>>> 11. Ke1 f3 >>>> >>>>Seems that crafty isn't doing any progress and score is totally wrong(seems to >>>>be draw). >>>> >>>>Is there an easy way to fix this? >>>>Or is this some kind off odd situation that is so rare that it is not >>>>worth fixing? >>>> >>>>Jarkko >>> >>> >>>That can be fixed. It is the classic problem of two connected passed pawns >>>being valued too high in a king and pawn ending. If they were separated, >>>white would win easily. But together, they can be stopped. I will look >>>at this and fix it. The code is already there, it just has a hole that I >>>apparently overlooked. >> >> These kind of position is very tricky. In some cases, white can forget about >>the black pawns and go with the king to help his pawns. AFAIK it can only be >>safely solved by search. >> >> José C. > > >I agree. But the way it is done, the error rate is higher than it should >be, I've explained why previously. Two connected passers are not what you >want in a king and pawn ending if your opponent has two passers that are >separated. I have code to do this right, but it has a hole in it that I had >overlooked...
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.