YouRang: White fell into a rather beginner trap on his 9th move:
click for larger view
He should have played 9.Be3, guarding the N and developing his DSB. Instead, he played <9.Nxe6?> allowing the face slap: <9...Bxf2+! 10.Ke2> (note: 10.Kxf2 Qxd1 ) <10...Qxd1 11.Kxd1 fxe6> and so white gives back the pawn, loses castling rights, and is stuck lagging in development. ~~~~
A few moves later, instead of moving his DSB to a safe square, black played <14...Rf8?!>, leaving white facing this position:
click for larger view
Here, white played <15.Be3?>. Why not take the bishop? Evidently, white feared 15.Kxf2 Ne4++ 16.Ke3 Nf2 forking R+B. But instead of 16.Ke3, white could improve with:
16.Ke1! <leaving e3 open for his bishop> Nf2 17.Nc3! <developing while clearing back rank) Nxh1 18.Be3 <sealing off f2, threatening Ke2 or Bg2 & winning N giving white 2B for a R> Nd4 <prevents Ke2 & threatens Nc2+, forking K+R> 19.Rc1 <not 19.Bxd4 Rxd4 allowing N to escape via f2> Rf3 <attack B> 20.Bxd4 <no choice anymore> Rxd4 21.Rc2! <now R covers f2, restoring Bg2 threat, and N will fall; it may continue 21...Nxg3 22.hxg3 Rxg3 23.Bxe6> with probable draw ~~~~
The game ended with <17.Na3?>, oblivious of the knight attack to follow: <17...Nfg4+!> [diagram]
click for larger view
The white K, restricted to the e-file by black's rooks, must accept a fork or be mated. For example, if 18.Bxg4 Nxg4+ 19.Ke2 (avoiding 19.Ke4 Nf2+) Rf2+ 20.Ke1 Rd2 and white has no good way to avoid ...Ne5 & Nd3#. In the game, white opted for <18.Ke4> hoping to win 2N for the rook <18...Nd3!> threatening ...Ndf2# <19.Bxg4 Nf2+> and white resigned seeing that he must now lose his bishop and rook, e.g. 20.Ke3 Nxg4+ 21.Ke4 Nf2+ forking K+R (note: not 21.Ke2 Rf2+ with same mating plan as above). | 
Kibitzing
sac: 18...Nd3
