My thoughts on the opening and early middle game:

Black plays an interesting gambit idea with 10...a4, which lures the white knight on c3 away from the center just long enough for Black to obtain a slight initiative.

Houdini plays

11.Ng5

- an active move, but Black counters with ...c5.

Black is playing very optimistically. He is trying to not only prevent White's expansion on the queen side (before launching into the typical dutch-king side attack), but he's even trying to get an edge in space there, before moving on the the king side. Not to mention the white bishop on e3 is lacking space (now that d4 is covered) and looks like a tactical liability in the making.

White therefore chooses to win the a4 pawn by forcing black to interfere with his queen's protection of the a4 square by playing

12.dc bc 13.Nxa4

Black plays 13...Rb8 to get off the long diagonal and pressure b2. Houdini gets his rook off the long diagonal too with 14.Rad1 but now black gets the initiative and expands on the king side.

14...h6 is a forcing move driving the knight to a bad square.

15...c5 - taking d4 again but with tempo this time thanks to the undefended knight on a4. This is where black starts cashing in on the pawn sac. The knight is attacked and must retreat.

16...g5 another forcing move. It looks like black threatens something like ...f4 (bishop is trapped because d4 was covered with tempo) and if white plays gf then ...g4 wins the Nh3. Not to mention ...g5 keeps the Nh3 from going to f4 and becoming active again.

It looks like White's best move is the move Houdini played 17.f4 but after ...g4 18.Nf2 h5 White has a bad knight on f2 and a cramped bishop on e3. Whites pieces lack harmony at the moment.

That is why after 19.h4 black doesn't play ...gxh3 (in order to get his own knight to g4) because that would allow White to fix his bad knight problem. It could go to g5.