F. Charly wrote:
Hi!
> Do you know this following position?
> 3b3r/4k3/rp3p1p/pPp1pPpP/P1PpP1P1/3P4/1K6/8 w - - 0 1
>
> HIARCS, at first, wants to take the "free" rook and
> suicide. Any human will only move it's king and it's a
> draw. After a short time HIARCS will finally choose 1.Kc2
> and the score will be -24.06 !!!
Seems that Shredder does a little better here ASAP:
1 33 -0.29 1.Ka2 Kd6 2.Kb1 Kd7 3.Ka1 Kd6 4.Kb1 Kd7 5.Ka1 Kd6 6.Kb1
2 33 -0.29 1.Ka1 Kd6 2.Kb1 Kd7 3.Ka1 Kd6 4.Kb1 Kd7 5.Ka1 Kd6
3 33 -0.29 1.Kb1 Kd7 2.Ka1 Kd6 3.Kb1 Kd7 4.Ka1 Kd6 5.Kb1 Kd7
It never even considers taking the rook.
Funny enough, all(!) other engines I have (including Rybka)
fall for bxa6 at a first glance. I did not let them
calculate too long, but evaluation of the kings moves are
that bad compared to taking the Rook, that it will take a
while, if ever, till they make it to the top.
> Morality: don't rely too much on computers assessments :-)
Sure not. I just found it a very interesting observation,
that in the example I gave the evaluation of the initial
position changes by 180 degree if the engine knows the end
of the game first. Simply as there're some moves that are
weeded out if it has to find them itself.
The reason to give this to the list was, that there was an
argument a while ago if it makes any sense to run analysis
backwards as this is technically a pretty complex thing to
do. Now, sometimes you do not have a nice example at hand,
but it will come along sooner or later. It took a while for
this to come along. (And I also like this forced under
promotions to Knights. As usual, the cavallry is always to
late ;)
Now, I did not try it but I'm pretty sure that if I take a
night or two to analyse the critical positions in the
example I gave one or the other engine will find the correct
move and evaluate accordingly. Still, in case of
autoannotating/blunder checking a finished game you usually
set the time per position to some amount more in the area of
seconds, or minutes and not days. That's why I did not use
to long calculation times in my test and it showed (IMHO)
in this more practical conditions that it would indeed make
quite some sense to run from the end to move 1.
--
Kind regards, / War is Peace.
| Freedom is Slavery.
Alexander Wagner | Ignorance is Strength.
|
| Theory : G. Orwell, "1984"
/ In practice: USA, since 2001
------------------------------------------------------------------------------
This SF.net email is sponsored by:
High Quality Requirements in a Collaborative Environment.
Download a free trial of Rational Requirements Composer Now!
http://p.sf.net/sfu/www-ibm-com
_______________________________________________
Scid-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scid-users
