It is a play on the game hacky sack:

Here's the position before that move.

click for larger view

Black is a pawn up but aside from the direct threat of 17.Qxe5, White is also threatening 17.Bxc6 Nxc6 18.Qxg7. What are the alternatives to 16...f6?

(A) <16...Qf6> defends both N and g-pawn but quickly loses to <17.Bxc6 Nxc6 18.Qd7+> so we can rule that one out straight away.

(B) <16...Qc7> loses the g-pawn to <17.Bxc6 Nxc6 18.Qxg7> but to add to that Black's K is stuck in the centre and his pawns are weak targets. 18...Qf4+ 19.Kb1 achieves nothing and if Black offers to trade Qs with ...Qe5 (with or without 18...Qf4+) then White's Q will take up a commanding position on g4. I reckon De Chen saw this and quite rightly rejected 16...Qc7 on that basis.

(C) <16...Qg5+ 17.Kb1> looks at first as if it solves both problems but it doesn't. What's more 18.Bxc6 Nxc6 19.Qd7+ is once again threatened. Black can get his K away from the centre with <17...O-O> but then <18.h4> destroys the illusion that the N is safe. Wherever Black's Q goes she'll be harried again until she gives up the N's protection.

(D) Finally we have the SF annotation's <16...Bxe4>. Once again this loses the g-pawn, this time to <17.Qxe5 Bg6 18.Qxg7> and, just as in (B), Black's K is stuck in the centre. There are some crucial differences from (B) though: With the B on g6 those pawns aren't so weak, White's Q is more restricted and <18...Qg5+> can no longer be answered with 19.Kb1. After <19.Rd2 Ke7> we would have a dynamic position with chances on both sides ... or White could opt for a draw with SF's <20.h4 Qe3 21.Rf1 Rhd8 22.Qf6+ Ke8 23.Qh8+>, etc..

So if De Chen did see the Q-sac coming then why did he prefer that over (D)? Maybe he just didn't like that position or maybe he saw all the way through to SF's draw and rejected it for that. Maybe he thought that, despite the isolated doubled pawns, his R pair would be more than a match for White's R,N and B.