The position is improbably sharp and at first glance hard to assess. White’s king is exposed, but his greater space and advanced kingside pawns seriously cramp Black, and represent a potential threat. One way or another, the active black pieces are insufficient to carry out an attack, and he needs to bring up his reserves. This is impossible to do without sacrifices, so...

If this move were not available, Black would have to play 24...Nf6!? 25.gxf6 Bxf6 26.Bd4 Rc8, retaining strong compensation for the pawn.

He loses immediately with 25.Bb6 Rxa3 26.bxa3 Nb5.

This attempt at counterplay fails, but going over to defence also loses:
A) 27.Bc5 Rxa3 28.bxa3 Rxc5 29.a4 Na3+ 30.Kb2 Rc2+ 31.Kxa3 Qb8! 32.Ra1 b2;
B) 27.Bd4 Rxa3 28.bxa3 Nxa3+ 29.Kb2 Nc4+ 30.Kc3 (30.Kxb3 Qa3+ 31.Kc2 Ne3+ 32.Kd2 Rc2+ 33.Ke1 Qb4+) 30...Qa2 31.Bc5 Qc2+ 32.Kd4 dxc5+ 33.Kd5 Ne3+ 34.Kd6 Nxd1.

White resigned.

The whole of Black’s position hangs by the knight on e5, which defends the f7- and g6-squares. Therefore White played...

exploiting the fact that the hanging Bb5 still serves a purpose – it covers the c6- and d7-squares, not allowing the other black knight to replace its colleague on e5. By contrast, the pointless check 14.0-0+ Kg8 would create a threat of 15...hxg5, and White’s attack fizzles out.

He also loses after 14...Rh7 15.Bxe5 dxe5 16.Rd1 Nd7 17.Bc4!. It turns out that the apparently lost white bishop comes into the attack with decisive effect.

Black resigned.

The pawn pair b7-c6 prevents White from developing an attack on the enemy king, which is stranded in the centre. The correct decision is the crude but effective: