[Retros] Re: e.p. problem

Noam Elkies elkies at math.harvard.edu
Fri Oct 8 18:00:53 EDT 2004


Yefim Treger writes:


> I have constructed a legal position (White move)

> with the following property:



> White has only one move and this move is en passant move!

> (what means that Black's last move was necessarily a pawn move

> to make this e.p. possible).



> In my position there are 16 pieces but I believe somebody

> will improve that...


In fact this has already been done before. In 1993 I constructed
a position such that White, in check, has only one legal move,
and that move is en passant -- and checkmate. Thus a computer
(or human analyst) not aware of retroanalytical issues would be
as far as possible from the correct evaluation of the position.
In fact, I constructed the position in the context of a rec.games.chess
contest for finding the worst evaluation by a given computer of a won
position (though I constructed it too late to participate in the contest).
I don't know whether such a position was already known then. Mine has
15 men, each a King, Rook, or pawn:

- - - - - - - -
- - - - - P - -
- - - - P R - p
- - - p R P p k
- - - p p K - p
r - - - - - - -
- - - - - - - -
- - - - - - - -

I wouldn't be surprised if 14 or fewer is possible.

--Noam D. Elkies




More information about the Retros mailing list