Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame: Best Move and Tuning

Author: Dann Corbit

Date: 13:22:46 03/06/06

Go up one level in this thread


On March 04, 2006 at 22:46:47, allan johnson wrote:

>On March 04, 2006 at 16:48:33, Dann Corbit wrote:
>
>>On March 04, 2006 at 16:30:09, Charles Roberson wrote:
>>
>>>
>>>
>>>    Telepath drew a won endgame in a blitz match this week. The position is:
>>>     8/8/8/5BB1/8/3p1kpp/7P/7K w - - 0 76
>>>
>>>    It only searched 6 ply and played Bxh3.
>>>    At 10 ply it played Bxd3.
>>>
>>>    Seems to me the better move is Bxd3 or h2xg3.
>>>    I've tried upping the outside passer value by 4x and adjusting
>>>     the code to call the d3 pawn an outside passer. Now, it makes
>>>     Bxd3 at 9 ply and h2xg3 at 10 ply then Bxd3 at 11 ply.
>>>
>>>    Seems I should be able to adjust it to see this earlier or would
>>>    that be overtuning?
>>
>>Rybka likes Kg1 (which wins, but is a very puzzling choice because hxg3
>>obviously wins very easily.)  Considering Rybka's long struggle to get off of
>>the drawing Bxd3 I would not be too worried about it if my program struggled
>>with it a while also.  However, a 6 ply depth is a great concern if your program
>>has more than one second of think time.  Why so shallow?  Severe time pressure?
>>
>>Analysis from e:\ttt.epd
>>3/4/2006 1:34:36 PM Level: 999 Seconds
>>Analyzing engine: Rybka v1.01 Preview 2.w32
>>
>>1) 0 76
>>    Avoid move:
>>    Best move (Rybka v1.01 Preview 2.w32): Kh1-g1
>
>
>Dann I'm puzzled.Your  Rbyka programme does not consider hxg3 [21#] in 00.08.56
>whereas Mike Hurd version announces mate in 21 in 4 seconds.Is this because of
>computer hardware or tablebases or is it something else I'm missing?
>Al

Not sure.  hxg3 seems a real no-brainer to me (as a human it is what I would
play and I don't think you even have to think to hard about it because it is
obvious that it wins -- the other two bishop pawn-capture moves are much more
complicated in my opinion, but I both of them win also)

In any case, I think that Rybka is supposed to use EGTB now, and since a UCI
engine, it should have the paths published to it.

So I don't understand why Rybka makes such a strange choice.  The other engines
I tried since then do not make that same choice and most of them make hxg3 right
away.



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.