Here is another application of the same idea: thinking about whether you can distract a guard by making a threat from a square it protects. Here White’s knight is on a dark square. So are Black’s king and queen, and they can be forked at e6. The trouble is that the square is protected by the Black bishop on f7. The bishop can't be captured; none of White’s pieces attacks it. Nor is the bishop quite protecting any pieces that White can take. It guards the d5 pawn, but so does the c6 pawn. But now you also can ask whether the restricted bishop is protecting any squares that White can occupy. Mostly you want to see if it is guarding any squares next to Black’s king, because then you may be able to draw it away by threatening mate on one of those squares. Think of it this way: the bishop on f7 is confined (if it wants to prevent the fork Ne6) to the diagonal running from a2-g8; it really cannot afford to go off on the e8-h5 diagonal—so look for play there.

What you find is that if White’s rook moves forward a square to h5, it checks the Black king. The king wouldn’t be able to capture it because the square also is attacked by a knight, so Black would have to respond to the check in some other way. Let's consider the possibilities:

One option would be for him to take the rook with his bishop, which would clear the way for the fork White seeks. The idea is 1. Rh5+, BxR; 2. Ne6+, Kf6; 3. NxQ. This looks strong, and it is—but take a moment to ask how Black will reply at the end. He has Rc8, attacking the knight; and this turns out to be surprisingly troublesome, since the knight has nowhere safe to go. Na6 is met with Ra8, renewing the attack and the knight still has no retreat. But instead of moving to Na6 in the first place, the knight can take a pawn on the way down by replying to Rc8 with Nc7xd5; and then after Black recaptures (c6xNd5), White grabs yet another Black pawn, and creates a passed pawn of his own, with Bxb5. So in the end White ends up winning a queen and two pawns for a knight and a rook.

Another possibility is that Black could reply to Rh5 by moving his king to f6. So look for your next check and find Rxf5. When the king moves again in response, you ask whether any Black pieces will have been loosened―i.e., left unprotected―by all this activity, and you find KxN for White.

This problem is a good example of a valuable general point: effective combinations often require both (a) pattern recognition, and (b) skill at identifying forcing moves—especially checks—and their consequences. The key move here—Rh5+—might have been spotted in an effort to make the fork work, or it might as well have been spotted just by examining the consequences of any checks you can give and any vulnerabilities of the enemy king. You notice that one of the possible Black responses to this particular check would be BxR; you imagine the board with your rook and his bishop moved, and a simple color scan or inspection of knight moves reveals the fork waiting at e6. If Black instead responds to the check with Kf6, you look for the next check, and so forth.