The wonderful thing about many endgame studies is that you can solve them a move at a time. The composer started with an idea, then worked backwards to reach the starting position. The solver need only retrace his footsteps. Try this position by Alexei Troitsky. It is White to play and win.

A piece behind, he clearly needs to create some strong threats. 1.Kxd6 is tempting, but leads to nothing after 1...Bf5 2.Ke5 Bb1 3.Rd6+ Kb7 4.Rd7+ Kc6 5.h7 Rg5+. No, the right move has to be 1.h7 - which also has the benefit of being easiest to analyse. To stop the pawn Black must play 1...Rg5+ 2.Kxd6 Rxh5, and what now? Well if we cannot queen the pawn, it will have to be mate, so 3.Kc7! forces 3...Be6 to stop Ra2 mate. Now Black cannot be allowed to take on h7 with check, so 4.Kb8! is the only consistent try. Black can only prevent Rd6 mate by playing 4...Bd5, and now 5.Rxd5 Rxd5 6.h8=Q wins.

Hang on. It doesn't win. 6...Rd8+! 7.Qxd8 is stalemate. So what's the answer? Take back 6.h8=Q and play 6.h8=R!! Black must reply 6...Rd6 to stop Rh6, when 7.Kc7! forces mate or win of the rook.