Michael Keller has been one of the finest composers of three-movers for many decades. This problem features some very subtle line play.
In the diagram three pieces guard d5. A Nowotny move to g4 will remove one of the guards, but immediate attempts to play a Nowotny fail simply to 1…Bxg3.
The key 1.a7 threatens 2.a8Q and 3.Qc6.
The a3 knight can defend by moving to c2 or c4, aiming to attack c6, but in each case cuts the rook guard of c5, leaving the f2 bishop to defend against Bxc5. It also leaves the knight in a position to prevent one of the Nowotny threats while closing the d1 rook’s guard of d5.
After 1…Sc4 2.Rg4 threatens 3.Se4 and 3.Sf5, with 2…Bxg3 3.Bxc5. 2.Bg4 fails because 2…Sd2 directly prevents 3.Se4 and White cannot play 3.Sf5 because it cuts the remaining guard of d5.
Similarly, after 1…Sc2 2.Bg4 is the correct continuation. 2.Rg4 fails because 2…Sd4 directly prevents 3.Sf5 while 3.Se4 would cut the bishop guard of d5.
There is a by-play variation 1…Bxe6 2.Rg6 threats 3.Sf5 and Rxe6; 2…Sf6, Sg7 3.Bxc7.
Altogether an impressive study in the closure of lines.
David Friedgood: A wonderfully intricate symphony by the typically strategic Keller 3-mover supremo. The by-play is a necessary construct ingeniously re-using the main-play pieces Bb6 and Rg5. Perhaps the judge disliked the humdrum key, but after that the solver will require accuracy to deal with the Nowotny.