OpenChess

TCEC Stage 1 - Season 3

http://www.tcec-chess.org/stage_1.php

STARTS FRIDAY APRIL 29TH @ 19:00 CEST

Official discussion thread: http://www.tcec-chess.org/viewtopic.php?f=23&t=127
NB: You need to be registered in order to participate or view the discussion on the official site.

For live chat please visit http://chessbomb.com/
NB: Don't pay too much attention to their built-in computer analysis since it's running multi pv which slows it down considerably.
NB: Engine updates are allowed only until the tournament starts,
and then the engine version numbers will be added.


Info:
CPU: Intel Core i7 980x @ 4198 MHz
MB: Asus Rampage III Gene
RAM: 12 GB Corsair
OS: Windows 7 Professional 64-bit
Threads: Up to 6 per engine.
Tournament length: 112 games.
Tournament type: Swiss (starts with the lowest boards each round)
Openings: Randomly selected, per game, from the new TCEC main opening book that currently consists of 468 openings.
This number will be increased later.
All openings are fixed to 10 moves and analyzed by Rybka & Houdini on the TCEC computer for 4 minutes per opening.


Time Control:
90 minutes + 30 seconds added per move for the whole game.


Graphical User Interface:
Type: ChessGUI
Resign: +/- 7.50 eval for 5 moves in a row (both engines)
Draw: An eval of +0.05/-0.05 for 5 moves in a row (both engines - only possible at move 30 or later)
If there is a pawn advance, or a capture by any kind, this draw counter will reset
Ponder: Off
Hash: Max 4096 MB per engine
Endgame tablebases: For any engine using these, all 3-4-5 men are available (Nalimov, Gaviota etc.)
Endgame cache: Up to 256 MB

For complete information on the rules and hardware, see the Information page.


Misc:
Stage 1 kicks off the new season. The top 16 finishers will move on
to stage 2 while the bottom 16 are out of TCEC for this season.

Stage 2 and beyond use the regular TCEC time control of 150' + 30".

Ratings are also implemented now, they will change as the
tournament progresses. The initial ratings are mostly based
on the CEGT 40/20 and CCRL 40/40 lists single CPU lists.

Complete season 1 archive: http://www.tcec-chess.org/archive.php
Complete season 2 archive: http://www.tcec-chess.org/archive.php

For news and updates check the News Forum regularly.

Now live!

http://www.tcec-chess.org/stage_1.php

Current World Computer Software Champion Shredder is the first to go 2/2:[Event "TCEC III Stage 1"][White "Shredder"][Black "Critter"][Round "2.7"][Result "1-0"]
1. Nf3 Nf6 2. c4 e6 3. Nc3 d5 4. d4 c6 5. Bg5 h6 6. Bh4 dxc4 7. e4 g5 8. Bg3 b5 9. Be2 Bb7 10. e5 Nh5 11. O-O Nd7 12. Qc2 Nxg3 13. hxg3 Qc7 14. a4 a6 15. Nd2 Be7 16. Bh5 O-O 17. f4 gxf4 18. gxf4 Rad8 19. f5 Qb6 20. Ne2 Nxe5 21. Qe4 f6 22. Qe3 exf5 23. Rxf5 Bc8 24. Rf2 Kh7 25. Qe4+ Kh8 26. Qe3 Kg7 27. Raf1 Qa7 28. Ne4 f5 29. Qg3+ Kh7 30. Qxe5 Rd5 31. Qf4 Rg8 32. Bf3 Rg6 33. Qh2 Rd8 34. Bh5 Rg7 35. Qe5 Kh8 36. N4g3 Bd6 37. Qf6 Qe7 38. Qxe7 Bxe7 39. Nxf5 Bxf5 40. Rxf5 Rd6 41. R1f3 Kh7 42. Kf2 Bg5 43. Rf7 Bd2 44. Rxg7+ Kxg7 45. Rf7+ Kg8 46. Ra7 a5 47. Ra8+ Kh7 48. Bf3 Re6 49. Ra7+ Kg6 50. Ra6 Kf6 51. Bg4 Re8 52. Rxc6+ Kg5 53. Bd7 Rf8+ 54. Kg3 bxa4 55. Rxc4 Rf1 56. Rc5+ Kg6 57. Bxa4 Kg7 58. Rc7+ Rf7 59. Rxf7+ Kxf7 60. Kf3 Bg5 61. Ke4 Be7 62. d5 Bd6 63. Nd4 h5 1-0Around move 25, both were initially indicating Qe4/Qe3/Qg3 with a draw, but then White playing 27. Raf1 seemed to turn the tide. Shredder had expected Qa7 in return with Ne4 to follow and a 0.02 score (upon more than 3min of thinking), but after Qa7 was indeed played by Critter, the score jumped dramatically. Who says there's no horizon effect at LTC?

A lot of excitement from sharp play in this match-up, but now it seems that IvanHoe will also go 2/2:
[Round "2.3"][Event "TCEC III Stage 1"][White "Protector"][Black "IvanHoe"][Result "*"]
1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 Nbd7 7. f4 Qc7 8. Qe2 e5 9. Nf5 h6 10. Bxf6 Nxf6 11. O-O-O exf4 12. Nd5 Nxd5 13. exd5+ Kd8 14. Qe4 Qd7 15. g4 h5 16. Qxf4 hxg4 17. Ne3 f5 18. Kb1 g5 19. Qxg5+ Qe7 20. Qf4 Qe5 21. Qf2 g3 22. Qd2 Rxh2 23. Qa5+ Ke8 24. Bd3 Rxh1 25. Rxh1 Be7 26. Nc4 Qf6 27. Qc7 f4 28. Re1 Kf7 29. Be4 Bg4 *It seems that moves 26 and 27 were White's main miscues. There might have been some lines with 29. Nb6 and winning the trapped rook, where either Black responds by promoting a pawn, and/or the ensuing Na8 is not realised as trapped until beyond the horizon.

In the end, the TCEC Rules had ChessGUI award the win to IvanHoe. It is still unclear whether it is really a win. The SF mainline had 5 queens on the board at various points.
[White "Protector"][Black "IvanHoe"][Event "TCEC III Stage 1"][Result "0-1?!?"][Round "2.3"]1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 Nbd7 7. f4 Qc7 8. Qe2 e5 9. Nf5 h6 10. Bxf6 Nxf6 11. O-O-O exf4 12. Nd5 Nxd5 13. exd5+ Kd8 14. Qe4 Qd7 15. g4 h5 16. Qxf4 hxg4 17. Ne3 f5 18. Kb1 g5 19. Qxg5+ Qe7 20. Qf4 Qe5 21. Qf2 g3 22. Qd2 Rxh2 23. Qa5+ Ke8 24. Bd3 Rxh1 25. Rxh1 Be7 26. Nc4 Qf6 27. Qc7 f4 28. Re1 Kf7 29. Be4 Bg4 30. Nd2 Re8 31. Bf3 Bxf3 32. Nxf3 Qf5 33. Qxb7 Qg4 34. Qb3 g2 35. a3 Bf6 36. Rxe8 Kxe8 37. Ka2 Be5 38. Nxe5 dxe5 39. Qb8+ Kf7 40. Qc7+ Kf6 41. Qd8+ Kg7 42. Qe7+ Kh6 43. Qf6+ Qg6 44. Qh4+ Kg7 45. Qe7+ Kg8 46. Qd8+ Kh7 47. Qd7+ Kh6 48. Qh3+ Kg7 49. Qd7+ Kf6 50. Qe6+ Kg5 51. Qxe5+ Kg4 0-1The SF mainline at the end continues: 52. Qd4 Kh3 53. Qh8+ Kg3 54. Qd4 f3 55. Qe5+ Kf2 56. Qd4+ Ke2 57. Qc4+ Ke3 58. Qc3+ Kf4 59. Qd4+ Qe4 60. Qc5 Kg3 61. d6 f2 62. Qg5+ Kf3 63. Qh5+ Ke3 64. Qg5+ Ke2 65. Qh5+ Ke1. Maybe the "auto-resign" feature should be turned off when the resigner has a queen (or 2 rooks). Perhaps someone with a database could say more, but I think "crazy pieces" (leading to stalemate) and unseen perpetuals mean that a score of -8 "could be" wrong about once every 2500 games or so, and maybe actually really affects only 25% of those.

Martin’s tournament computer must have one of the best cost/usage ratios around; I’d hate to see his electricity bill. And congratulations to Martin in increasing recognition and consultations on ChessVibes!

MoldyJacket wrote:Martin’s tournament computer must have one of the best cost/usage ratios around; I’d hate to see his electricity bill. And congratulations to Martin in increasing recognition and consultations on ChessVibes!

Thanks.

If you enjoy TCEC there is a donate button on my info page. The money will be spent on buying commercial engines and paying the power bills.

Best,
Martin

A tremendous slugfest from the match-up we've come to expect this from, with the evals (and proposed plans) in a material imbalance bouncing to and fro. Stockfish never particularly agreed with either (though over time tending more toward Black), and the ChessBomb discussion was unclear to boot, though most seemed to prefer White for much of it (the perception that Houdini was overevaluating the queen was a common topic).
[Event "TCEC III Stage 1"][White "Rybka 4.1"][Black "Houdini 1.5a"][Round "3.2"][Result "0-1"]
1. d4 d5 2. c4 c6 3. Nc3 Nf6 4. Nf3 a6 5. a4 e6 6. Bg5 Nbd7 7. a5 dxc4 8. e4 h6 9. Bxf6 Nxf6 10. Qa4 Bd7 11. Qxc4 Rc8 12. Be2 Be7 13. O-O c5 14. e5 Nd5 15. Nxd5 Bb5 16. Qxb5+ axb5 17. Bxb5+ Kf8 18. Nb6 g5 19. Nxc8 Qxc8 20. Rfc1 c4 21. Bxc4 Kg7 22. Ne1 h5 23. Be2 Qd7 24. Rd1 Qd5 25. Nc2 Qb3 26. Ne3 g4 27. Nc4 Qb5 28. Rac1 Bb4 29. h3 gxh3 30. Rd3 h4 31. Rxh3 Bxa5 32. Rf3 Rf8 33. Bf1 Bd8 34. Rcc3 Kg6 35. g3 Qb4 36. Rb3 Qe1 37. gxh4 Kh7 38. Ne3 b6 39. Rf4 f5 40. Kh2 Rg8 41. d5 Bc7 42. d6 Bd8 43. Rbb4 Qa1 44. Rb5 Rg6 45. Kh3 Qc1 46. Bg2 Kg7 47. Kh2 Kg8 48. Kh3 Kf7 49. Bf1 Rh6 50. Kg3 Qc6 51. Nxf5 exf5 52. Rxf5+ Kg6 53. Rf4 Qh1 54. Rg4+ Kf7 55. Bc4+ Ke8 56. h5 Rh8 57. Rf4 Rh7 58. Bd3 Qg1+ 59. Kf3 Rxh5 60. d7+ Ke7 61. Bc4 Qd1+ 62. Ke4 Qc2+ 63. Kd4 Rh7 64. b3 Kxd7 65. Rf7+ Rxf7 66. e6+ Ke7 67. exf7 Qxf2+ 68. Ke4 Qg2+ 69. Kd3 Qg3+ 70. Kd4 Bc7 71. Ke4 Qf4+ 72. Kd3 0-1Both sides thought they were winning by about 0.25 for some time, then Rybka got a bit excited with the d6 push on moves 41-42, which then immediately was seen as evaporating any advantage (it seems that White's pieces were all optimal placed, with nothing more to do, as it were). Houdini then suddenly spiked to -0.98 with Qc1, though it wasn't immediately obvious how to provoke enough weaknesses to beat off White's coordination, and dropped to merely -0.69 after a few moves. Rybka stayed at 0.00 for some time, and at this point one question was whether Black could exploit some sort of zugzwang to force a weakening (recall that Houdini has much better mid-game zugzwang recognition than Rybka). After some King zig-zagging (Kg7-h7-g8-f7), it seems that this occurred upon White's 49. Bf1 (rather than a King shuffle), and within a move Rybka had agreed Black was winning (-0.60), and with Houdini claiming -2.21 the advantage augmented rather quickly. White gave up the Knight for two pawns to yield RRB+5 versus QRB+1, and the heavy material then won the day, with Rybka realising it was -2 upon 54. Rg4+ (the first "real think" [longer than 15s] in a couple of moves). White's King had no home, and it was over rather quickly (though the mop-up was not the swiftest, the result seemed pretty inevitable, with there really no hope for a fortress draw unless Black rashly exchanged down to a drawn QP vs RP(P) or something --- contrarily, keeping the opposite coloured bishops helped the stronger side). There was again an outside question of whether Black could break the fortress at the time of resignation as SF@30s didn't see much progress, though in this case Rybka's score had dropped to -11, and presumably there was some material loss in such a PV.

Thank you BB+ for your very interesting game comments.

After 3 of 7 rounds, Houdini is the only engine perfect so far.

Critter is having a horrible tournament and is in place 28 with 0,5 of 3.
Currently it is playing vs Deuterium in round 4 and needs a win in order
to have a shot at qualifying for stage 2.

Table:

Round 4 saw Houdini and Shredder win against IvanHoe and Komodo respectively. The latter game was rather nice, IMO.
[Event "TCEC 3 Stage 1"][Round "4.2"][White "Shredder 12.0"][Black "Komodo 4165"][Result "1-0"]1. c4 e6 2. Nf3 d5 3. g3 Nf6 4. Bg2 Be7 5. O-O O-O 6. b3 d4 7. e3 c5 8. exd4 cxd4 9. d3 Nc6 10. Qe2 a5 11. Ne5 Nxe5 12. Qxe5 Nd7 13. Qe2 Qb6 14. Na3 Nb8 15. Qe4 Rd8 16. Nb5 Na6 17. Rb1 Nc5 18. Qe2 Na6 19. Qf3 Bf6 20. Re1 Be7 21. Bb2 Bf6 22. a4 e5 23. g4 Nb4 24. Qg3 h6 25. h4 g5 26. hxg5 hxg5 27. Bc1 Kg7 28. Bd2 Nc6 29. Bd5 Ne7 30. Rxe5 Bxe5 31. Qxe5+ Qf6 32. Qxf6+ Kxf6 33. Be4 Nc6 34. f3 Be6 35. Kg2 Rac8 36. Bxc6 bxc6 37. Na7 Rc7 38. Bxa5 Rdd7 39. Bxc7 Rxc7 40. Nxc6 Rxc6 41. b4 Bc8 42. Rh1 Ra6 43. Rh8 Bb7 44. a5 Ke5 45. Rb8 Ba8 46. Kg3 Ke6 47. f4 f6 48. Rh8 gxf4+ 49. Kxf4 Kd7 50. Rh7+ Ke6 51. Rh6 Kd7 52. Kf5 Bb7 53. g5 fxg5 54. Rxa6 Bxa6 55. b5 Bb7 56. a6 Bg2 57. Kxg5 Kd6 58. b6 Kc6 59. b7 Kc7 60. Kf4 Bc6 1-0
How many humans would play Rxe5 at move 30, sacrificing the exchange for a pawn, to be directly followed by a queen exchange? Admittedly, White's pieces are way better, but there's no clear path to an "obvious" advantage from what I can see. Even at move 41 and beyond, I would not personally have been too sure of White's winning chances, but the mass of passed pawns won the day in the end.