[Retros] 9 white pawns, cont'd

[Noam Elkies]

I wrote:


| > With 9 white pawns, there is e.g. the followoing joke problem

| > (but only the removal of some pawns allows a #1):

| > George Leathem

| > The Problemist FCS 821 05/1933

| > 8/3pPKPp/4N1N1/4PkP1/3P1p1P/3PRPRP/8/5Q2

| > #1



| Looks like each of White's 9 pawns can be removed: Qd3/h3/xf4,

| Nd4/g7 or Nh4/e7, and Re5/Rg5 would all be checkmate.  What I

| don't see is why the Black pawns on d7 and h7 are needed.


On further thought, even after removing one of White's nine pawns
the position is illegal WTM unless it's the e5 or g5 pawn that's removed
(when Black could have just played 0...e5xf4 or g5xf4); perhaps that's
what Mario R. was hinting at.  Removing Pe7 also results in an
impossible pawn configuration (where's the a-pawn?).  I haven't
tried to check whether any of the other eight choices give White
too many captures.  In any case the pawns on d7 and h7 don't help.

There's also a dual mate Qb1(d3) in the -Pd3 branch.  It seems that
this is easily fixed by moving Re3,g3 to e1,g1, which also lets us
eliminate the retro-illegality issue by moving wPf3 and bPf4 down to
f2 and f3 -- again assuming that we can retract 0 e2xd3++ (or 0 g2xh3++)
Kf5 without introducing too many pawn captures on any of the seven
branches that keep Pd3 and Ph3 on the board.

NDE