Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame: Best Move and Tuning

Author: Uri Blass

Date: 19:10:18 03/04/06

Go up one level in this thread


On March 04, 2006 at 21:02:37, Charles Roberson wrote:

>On March 04, 2006 at 20:28:34, Uri Blass wrote:
>
>>On March 04, 2006 at 18:10:27, Charles Roberson wrote:
>>
>>>
>>>  Yes, in following the match after Bxh3 black makes g3xh2.
>>>  Eventually the white king takes h2 and we have a 5 man endgame.
>>>  The Nalimov tablebase server sees it as a mate in 21 against black.
>>>
>>>  However, at blitz time controls (without EGTB's) a mate in 21 is too
>>>  much to see at 1 sec per move which is what I end up with on the 76th
>>>  move of a G/5 match.
>>>
>>>   So, I like the much more "human" moves of Bxd3 and h2xg3.
>>>
>>>   Maybe the best idea is to forget it and put in EGTB's??????
>>>   Seems that would save me a fair amount of analysis down the road.
>>
>>I think that the best idea is simply to improve your human knowledge.
>>I also did not know it but I learned from the position.
>>
>>2 bishops are simply usually enough to stop a pawn in the 6th and no need for
>>the king to be close to the pawn for that target.
>>
>>Even the following position is a simple win for white.
>>
>>[D]K7/8/8/6B1/8/3p3B/4k3/8 b - - 0 2
>>
>>The main question is how could your program not win the game.
>>
>>Uri
>
>    At the end of a blitz game it was making only 6 ply searches in < 1 secs.
>    It fell into 3 fold rep of position after black plays d2d3.
>    I checked the Nalimove tablebases on the position you posted and its a
>    mate in 21. Seems a bit much for a program to see in < 1 secs per move.
>    The machine is a 1.6 GHz Athlon 32 bit cpu.
>
>    Also, for Telepath I have not put in transposition tables yet.

I think that tranposition tables are not needed and 6 ply search is enough to
get the white king closer to the pawn and in the same time to prevent the black
king to defend the queening square.

Note that I also think that programs should get more than 6 plies even in 0.5
seconds on your hardware even with no hash tables.

Movei with a single hash entry and with no tablebases can get easily more than
it
(I have no option to disable hash but single hash entry is really not
significant espacially when movei use hash only for order of moves).

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.