Hi guys and girls!

I have some food for thought for you! I got the chance to test the Bh3-idea in 2010 in a serious game (time control: 2h/40 moves, 30min for the rest of the game). The game was really complex and although I made a few inaccurate moves after the opening the game remained quite complex all the time. In the end I managed to win the ending. Theoretically speaken Black must only look at the improvement 25...Rd8, which is given as completely equal by Houdini 1.5.

Heiduczek, Josef - Lötscher Roland
(Böblinger Open, 27th of December 2010)

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. Be3 Bg7 7. f3 Nc6 8. Qd2 O-O 9. Bc4 Bd7 10. Bb3 Rc8 11. O-O-O Ne5 12. Kb1 Nc4 13. Bxc4 Rxc4 14. g4 b5 15. b3 b4 16. bxc4 bxc3 17. Qxc3 Qc7 18. g5 Nh5 19. Ka1 Bh3 20. Rhg1 (20. Rhe1 is White's main alternative. Something like 20...Rc8 21. Rb1 Bg2 22. Qd3 Bxf3 23. c3 Bg4 will follow; 20. Bf2 Nf4 is not what White should aim for.) 20... Be6 21. Rb1 d5 Now Black's idea of
inserting Bh3 and Rg1 becomes clear. White's h-pawn is hanging. Once the h-pawn is gone Black's knight enters the game via g3. 22. Qd2 A human move. White simply defends h2 and shuts the dragon bishop down by playing c3 soon. However Black can equalize completely now. (22. c5!? dxe4 23. fxe4 Qxh2 looks like the critical line) 22... dxe4 23. fxe4 Bxc4 24. Rb4 (The main
alternative consists in 24. c3 Rd8 25. Qc2 (25. Rgd1? e5 26. Nb5 Rxd2 27.
Nxc7 Rxa2#) (25. Rbd1 Rd6) 25... Bxa2!? (25... Nf4 26. Rgd1 Nd3 27. Rxd3 (or 27.h4 Ba6 28. Rxd3 Bxd3 29. Qxd3 e5) 27... Bxd3 28. Qxd3 e5 29. Rb2 exd4 30. cxd4 a5 also gives Black sufficient compensation.) 26. Kxa2 Bxd4 27. Bxd4 Rxd4 28.
Rb7 Qxb7 29. cxd4 Qa6+ 30. Kb1 Nf4 31. Rg3 Ne6 32. d5 Nd4 33. Qd3 Qb6+ 34. Kc1
a5 looks fine for Black.) 24... Ba6! (24... Rd8?! 25.Qc3) 25. c3 (25. Rd1
Rd8 26. Kb2 Qe5) 25... Be5?! I wanted to force White to move his h-pawn before
playing Rd8. Unfortunately it turned out that White does not have to care
about it. (25... Rd8! is more accurate. 26. Qc2 (26.Qb2 Nf4 27. Bxf4 Qxf4 28. Re1 Qxg5 29. e5 h5) 26... Nf4 27. Rd1 e5! when White should force the
draw by playing 28. Nb5 (28. Bxf4?! exf4 already looks slightly better for Black.) 28... Rxd1+ 29. Qxd1 Bxb5 30. Rxb5 Qxc3+ 31. Kb1 Qxe3 32. Rb8+ Bf8 33.Rxf8+ Kxf8 34. Qd8+ Kg7 35. Qf6+ Kg8 36. Qd8+) 26. Rgb1! Now White might be
slightly better. 26...Rd8 (26... Bxh2? 27. Ra4 Bc4 28. Nc2 Be5 29. Rxa7) 27. Ra4 Bc8? too passive. I should have played 27... Bc4 not
worrying about 28. Rbb4 Bf1. 28. Rbb4! a6 (28... Bd7 29. Rc4 Bxa4 30. Rxc7
Bxc7 is also clearly in White's favour.) 29. Rc4 Qb7 30. Rab4 Qa8 (30... Qxe4?? 31. Nf3! wins immediately.) 31. Qc2 Bb7 32. Nb3 Bc8? (32... Bc6 33. a4 Bb5 34. Rd4 Bxd4 35. cxd4 Bd7 was the last real chance for Black.) 33. Bb6
Rf8 34. Nc5? (34. Bc7! Bg7 35. Rb8 Qa7 36. Qd2 and Black's position is
very cramped.) 34... Nf4 Suddenly Black is back in the game. 35. Qd2 Ne6 36.Qd5!? White wants to simplify the position in bad time trouble. 36...Qxd5 37.exd5 Nxg5 38. Ra4 Nf3 39. Nxa6 Bxh2?! (39... Nd2! 40. Rc5 f5 41. Bc7 Bg7 and the chances are about equal in this complex ending.) 40. Nb4 Ne5?! (40...h5) 41. Re4! g5 42. Bc7?! (42. Bc5! Re8 43. Ra7 gives White the
advantage.) 42... f6 43. Nc6?! Bd7 44. Ra6?! (44. Nxe7+ Kf7 45. Ra6! g4 46. Bd6 g3 47. Bc5 looks quite unclear.) 44... Bxc6 45. dxc6 (45. Rxc6 Nxc6
46. Bxh2 f5! 47. Re1 Na5 48. Rxe7 f4 -+) 45... Rc8 46. Bxe5 Bxe5 have a
look at this dragon bishop (not to speak about Black's four potential queens)!
47. Kb2 h5 48. c7?! a desparate trap. (48. Kc2 h4 49. Re3 Kf7 should also
win for Black.) 48... Kf7 (Of course not 48... Rxc7?? 49. Rxe5 fxe5 50.Rg6+) 49. Rxe5 fxe5 The rest doesn't need explanations. 50. Rc6 h4 51. Kc2 e6
52. Kd3 h3 53. Rb6 g4 54. Rb8 Rxc7 55. Rh8 Kg6 56. Ke4 Rxc3 57. Kxe5 Kg5 58.
Rg8+ Kh4 59. Kf4 Rc4+ 60. Ke3 Kg3 61. a4 Rxa4 62. Rh8 h2 0-1

And what happens after 20.Rd2 ? I think objectively speaking White is better, but in practice Black probably enjoys an equal share of the spoils.

bragesjo wrote on 12/23/10 at 15:14:53:


That's right. Houdini takes Black's play in the dragon much more serious than other engines do. I recommend to create a database of critical dragon positions and let Houdini fight these positions out over night with another engine. In the Fritz GUI you can use the "engine match" option for that purpose.

Bye the way, I do now understand the advantages of 19...Bh3 over 19...Be6. Since 20...Bg2 is a threat White should move his rook from h1. The main idea is to remove the defense of the h-pawn (after 20.Rhg1 or 20.Rhe1). After a future d-pawn advance, the h-pawn becomes vulnerable and once the h-pawn is gone, the knight on h5 enters the game via g3. Moreover after the natural 20.Rg1 White has to be even more careful that Black's knight won't find its way to the e2-square (via f4 or g3). So the inclusion of 19...Bh3 20.Rhg1 before 20...Be6 seems to help Black in some critical variations. Every other 20th move of White gives Black new options, so I think that Houdini's number one choice of 19...Bh3 may well be the best move in the position.

bragesjo wrote on 12/21/10 at 14:18:10:


Hmm, Houdini 1.5 (the new beast, much stronger than Rybka 4) recommends 19...Bh3 instead of 19...Be6 (his second choice). The main idea is to catch the f3 pawn (via Bg2, Bxf3). So White should either move his bishop (allowing Nf4) or move his rook from h1. After the natural 20.Rhe1 Rc8 Houdini already thinks that Black is close to equal. Maybe 20.Rhg1 is better.
I don't really get the pattern, but playout results so far have been pretty equal. It seems worth to give it a closer look.

How is Black doing in the Burnett variation (well Yugoslav attack with 9.Bc4 and 12.Kb1 Nc4 13.Bxc4 Rxc4 14.g4 b5 15.b3 b4 line, let's include 16.bxc4 bxc3 17.Qxc3 Qc7)?
I remember vaguely that bragesjo has done some work on it, but unfortunately I cannot find it any more.

I have done some analysis, too. I was surprised to find that in most lines Black's chances seem pretty good. The only line which is really worrying me as a dragon player is the line starting with 18.g5 Nh5 19.Ka1. Has anyone found a way to keep chances more or less equal here?