Figure 3.3.5.2[White to move]

The pattern we are considering should be no less evident when the three pieces involved are arranged horizontally, as they are here: rook-knight-rook on the fourth rank. Another signal for White to see is that Black’s rook is loose; it would make an easy target if White could get his knight out of the way and give check at the same time. He can’t, for again his knight and the enemy king are on squares of different colors. But it still pays to study the king and its vulnerabilities. Black's king has little room to move. All its flight squares on the fourth rank are sealed off by White’s king and pawns; so is g5. Indeed, the only place it can go is g6. It follows that White would mate if he could safely attack the king’s current position and g6. A knight can do that sort of thing—but from what square? In theory there are two possibilities: e7 and h4. The latter square is out of reach, but by moving the knight to d5 White threatens Ne7#; the initial move to d5 therefore is as threatening as a check. Black has to fend off the mate threat. If he tries to create a flight square for his king by moving his knight, then of course White has RxR—the point all along.

Ah, but wait: we must consider whether Black might somehow move his target to safety and extinguish the mate threat. He can—sort of—with 1. …Rc7, avoiding White's threat of RxR and also guarding the mating square e7. But this still sacrifices the exchange, as White then has 2. NxR, NxN.