Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Help with strange tablebase output - Crafty-18.6 & Fritz6

Author: Uri Blass

Date: 13:30:36 03/16/01

Go up one level in this thread


On March 16, 2001 at 16:22:07, John Hatcher wrote:

>On March 16, 2001 at 16:09:53, Dann Corbit wrote:
>
>>On March 16, 2001 at 16:02:20, John Hatcher wrote:
>>
>>>Here is a position arising from Rueben Fine's "Basic Chess Endings", No. 309b -
>>>
>>>[D]r7/4P1k1/3K4/8/8/8/8/5R2 w
>>>
>>>White wins after 1. Ra1.  This solution was found instantly (of course) by both
>>>Wcrafty-18.6 (using Winboard) and Fritz6, both using the same 5-piece Nalimov
>>>tablebases.
>>>
>>>But now it gets strange.  What if White temporizes first with 1. Rg1+  Kf7,
>>>arising at this position:
>>>
>>>[D]r7/4Pk2/3K4/8/8/8/8/6R1 w
>>>
>>>Wcrafty-18.6 instantly finds the win after 2. Rf1+  Kg7 (not 2. ...Ke8? 3. Rf8#)
>>>3. Ra1, transposing into the first position.  However, Fritz6 doesn't see the
>>>win!?  It's engine output reads:
>>>
>>>Analysis by Fritz 6:
>>>
>>>1. = (0.03): 10.e8R Rxe8 11.Rf1+ Kg6 12.Kd5 Kg5 13.Rf7 Rd8+ 14.Ke5 Re8+
>>>2. = (0.00): 10.Re1
>>>3. = (0.00): 10.Rh1
>>>4. = (0.00): 10.Rd1
>>>5. = (0.00): 10.Ke5
>>>6. = (0.00): 10.Kd5
>>>
>>>How can Fritz6 see the win in the first position and not see it in the second?
>>>The most obvious move, 2. Rf1+, isn't even listed in the engine output!?
>>>
>>>What's going on here?  Can anyone duplicate this strange output?
>>
>>That is very odd.  Every engine I tested (which knows about tablebase files)
>>found it immediately.
>>
>>I have the 5 men files.  Maybe Fritz has only the 4 men files available?
>
>No, Fritz6 and Wcrafty-18.6 are accessing the same 5-piece tablebase files.
>
>Just to be certain, I run the native Chessbase version of Crafty-18.03 under
>Fritz6 and it instantly found the win in the second position after 2. Rf1+.
>
>Fritz6 accesses and uses the 5-piece tablebases just fine, as evidenced by the
>fact that it instantly found the win (1. Ra1) in the first position.  I've been
>using the 5-piece tablebases under Fritz for a while, and they've always worked
>just fine.
>
>Even stranger, in the first position, Fritz announces mate in 'x' moves if White
>plays 1. Rg1+ instead of 1. Ra1, but then "loses" the mate when it reaches the
>second position.   I'm stumped.
>
>JOHN

The reason is clear.

Fritz calls the tablebases everywhere except cases when the remaining depth is
small.

I guess that after 1.Rg1+ it can see the mate because it never reaches the
position after Kf7 Rf1+ Kg7(it probably looks at the tablebases when the
remaining depth is 3 so it does not have to search to depth 3)

After 1.Rg1+ Kf7 it cannot see the mate because remaining depth of 2 plies is
too small for it to look at the tablebases so it can see repetition and
evaluates the position as a draw.

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.