🚀 go-pugleaf

In article <1992Aug20.085014.29492@sniap.mchp.sni.de>, geissler@angelika.uucp (Norbert Geissler) writes:
|> In article <Bt932y.C64@ux1.cso.uiuc.edu> dwight@ux1.cso.uiuc.edu (Dwight Robert Stone) writes:
|> >nash@visus.com (Richard V. Nash) writes:
|> >
|> >>  +----+----+----+----+----+----+----+----+
|> >>8 |    |    |    |    |    |    |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>7 |    |    |    |    |    |    |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>6 | *P |    |    |    |    |    | *K | *P |
|> >>  +----+----+----+----+----+----+----+----+
|> >>5 |    |    |    |    |    |    |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>4 |  P |    | *B |    |    |    |    | *P |
|> >>  +----+----+----+----+----+----+----+----+
|> >>3 |  B |    |    |    |    |  K |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>2 |    |    | *P | *B |    |    |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>1 |    |    |    |    |    |    |    |    |
|> >>  +----+----+----+----+----+----+----+----+
|> >>    a    b    c    d    e    f    g    h    
|> >
|> >
|> >>This is the position just before c2c1B. My guess as to why GNUchess did this is  
|> >>because its evaluation function "reasoned" that it was better to go with 3  
|> >>bishops against 1 than 2 against 1. If the pawn would have promoted to a queen,  
|> >>no doubt it would have been taken by the bishop on a3. As the game went, the  
|> >>white bishop moved away.
|> >
|> >>Sound plausable? Sound like something a silly machine would do?
|> >
|> >	This seems more to me like something a silly machine would do.  When
|> >you are up by this many peices it is almost always beneficial to trade away
|> >peices.  Besides, if the bishop on a3 took the queen it would not be 2 against
|> >1, it would be 2 against 0 after the bishop at d2 took the bishop at a3 (now
|> >c1).  This would leave GNU chess with a white and black bishop and 3 pawns
|> >(with the pawn in row a ready for easy promotion after the removal of the
|> >opponent's pawn which would not be able to be prevented by the king (sitting
|> >all the way over in f column).  Besides, two bishops (of opposite colors) and
|> >a king vs. a king is a workable mate solution.
|> 
|> GNU CHESS is moving here exactly like would have moved in a real tournament,
|> perhaps I would have promoted even to a Knight, whoever the opponent would
|> have been. It's a great psychological move which tells the opponent that
|> it's time to resign. 
|> In my personal practice, it worked every time.
|> 

I don't understand the discussion about what move is best. Any position can
be classified as follows:

1. Position is won for white and white knows how to win (whatever his
   opponent does)
2. Position is won for white and white does not know how to win for all
   all possible moves of his opponent
3. Position is draw and both sides know how to avoid losing the game (whatever
   their opponents do)
4. Position is draw, but either (or both) sides does not know how to avoid
   losing the game for all possible moves of his opponent
5. like (1) for black
6. like (2) for black

The position discussed above is obviously of catagory (5), WHATEVER GNUCHESS
PROMOTES ITS PAWN TO. So all moves are (from objective standpoint) exactly
equal. And a "perfect" evaluation function would give exactly the same value.

If the original question was only concerned with what actually happened in
Gnuchess' evaluation function, I have to excuse for the misintrepretation.

BTW: What do you think: If a chess program has a perfect evaluation function
(returning 1 if it can force a win, -1 if its opponent can force a win and 0
if neither side can force a win), there is a set of best moves (those returning
the maximum value) for each position. Are all these moves really equally good?
I mean: if the return value for this set of moves is 1, they surely are equally
good, since all moves force a win. But in either of the two remaining cases
you could argue that the programs opponent could chose a bad move. So you could
evaluate your moves by the probabality that your opponent choses a bad move.
But how measure this propability?

Sorry if these points miss the intention of this group.
|> --- Norbert
|> Norbert Geissler, SIEMENS-NIXDORF AG, SNI AP 154, Munich, Germany
|> E-MAIL: norbert.geissler@sniap.mchp.sni.de   
|> ..for chess, that superb, cold, infinitely satisfying anodyne to life, I feel
|> the ardour of a lover, the humility of a disciple.  (Herbert Russel Wakefield)