lichess.org
Donate

Hardest/Impossible Composition Challenge?

#11
@demonolith

Underpromotion to make the opponent think whether it's even a good idea to capture the piece can be a trick to make him loose time, but if the correct move is to not take the underpromoted piece, you should have promoted to a Queen, right, because it seems your opponent will be better anyway. If the correct move is to take the underpromoted piece, it doesn't matter at all and you could have promoted to a queen.

That said, some funny examples here:
timkr.home.xs4all.nl/chess2/minor.htm

The Amelung-Erler position (K+pawn, K+pawn) probably comes closest to what you have in mind. - but still involves a stalemate.
#12
..another interesting find from Tim Krabbe's site:

Sokolsky - Ravinsky 1938


... also involving a stalemate, but indirectly. After queening, Black can keep checking the white king with the rook, as white can never capture it without stalemating. So technically queening a8=Q? could end in perpetual checks, not a stalemate.

Promoting to a bishop seems to be winning, but the question in this position is whether you should push the pawn at all... but that's way beyond me. (my friend Stockfish would play b4 instead of promoting immediately).

#13
@Panagrellus Thank you for the link, many interesting positions.

I really meant a case where psychology is not involved but I assume it is impossible. For example, white is winning but promoting to a queen is winning less quickly than promoting to a bishop, and the next move of the opponent has to be different depending on this. It is how I understood the challenge posed by the author of the discussion.
#14
@Demonolith

This would be a nice use case for Lichess's gigantic database of games.
database.lichess.org/

One could write a script to do the following:

1. In the games database, search for all underpromotions to R or B. Most of them will be jokes, I'm afraid.
2. Filter for underpromotions that are meaningful - e.g. keep games if a quick computer evaluation gives more than +2 difference in favor of R or B.
3. Filter for forced stalemates in case of =Q instead of=R/B

If any games remain after this procedure, they might be interesting positions. But intuitively I also feel that what @AyrtonTwigg is looking for is just impossible.

#15
@Panagrellus This is a great idea but I do not know how to do this. I have never used this database. It would probably be only a few lines of code for someone who knows how it works.
#16
I don't think that the database is searchable that way online. But i wonder what software could do that locally.

Among the software that can handle big databases or books, are there some that are scriptable?

I don't know how much in the script ideas is feasible. I would leave the shallow evaluation to the end of the filtering process, or restrict to engine analysed games in the database. The database can be downloaded on a monthly basis.
#17
@Demonolith

I wouldn't know how to do this search either... it was not a suggestion for you, just a random idea. And because of the size of the lichess database, I think it would not be that trivial.

I'm tagging @e4Guardian here, who would be someone who could do something like this - just in case he's interested in weird underpromotions....

#18
It turns out that Tim Krabbe did come up with a problem that fulfilled the requirements that @demonolith set out (at the time of its composition in the 1970s).

"CHALLENGE:
Create a chess problem (in regular chess), where one of the moves has to be an underpromotion to a Rook or Bishop, and the purpose of the underpromotion is not to avoid stalemate."

Mate in 2:
lichess.org/analysis/8/4P3/8/3p4/2p3p1/1pP2kPp/1P5P/R3K2R_w_-_-_0_1



(hint: you have to creatively apply a pre-1972 FIDE rulebook on regular chess. Stockfish can't help you with this one).

This topic has been archived and can no longer be replied to.