TCEC Stage 2a - Season 3 (150m + 30s) LIVE! Starts 18/5

[Sarciness]

I concur about the eval graphs- very nice addition but the length of the y-axis needs to be increased by, say, 2 or 3 times so that more detail can be seen.

Also, maybe cap eval at +/- 5 or 6 instead of 8 for easier-to-view fine detail?

[Martin Thoresen]

As already mentioned by BB+, an eval graph has now been added to the live page. You can hide it if you feel like it.
I will update the other pages with the same graph later today.

[BB+]

I think my suggestion for the eval graph was more along the lines of making a logarithmic(?) y-axis somehow. E.g., the behaviour between 4 and 8 is not as interesting as that between 0 and 4, and so it perhaps could be condensed. My guess is that finding the "optimal" eval graph params are about as difficult as the "optimal" draw/resign rules.

For instance, in the graph below, the 0.0 to 1.0 range is more interesting (to me at least) than the 2.0 to 7.5 range, even though most of the graph wiggles happen in the latter. If I find the time, I will re-process the data-points therein, to give an example of what I personally would prefer.

[BB+]

Battle of the leaders in Round 6 saw Stockfish prevail over Junior. White's nascent kingside pressure was rather suddenly winning (see the graph in the previous post), as Black's queenside invaders turned out to be much out of play.
[White "Stockfish 2.11"][Black "Junior 12.5"][Result "1-0"][Event "TCEC 3 Stage 2a"][Round "6.4"]
1. e4 c6 2. d4 d5 3. e5 Bf5 4. Nf3 e6 5. Be2 Nd7 6. O-O Ne7 7. Nbd2 h6 8. Re1 Rc8 9. Nf1 Bg6 10. Ne3 Nf5 11. a3 Be7 12. b4 a5 13. Bd3 Nxe3 14. Bxe3 Bxd3 15. Qxd3 b5 16. Bd2 axb4 17. axb4 Nb6 18. c3 Ra8 19. Bf4 O-O 20. Nd2 Ra4 21. Qg3 Kh8 22. Qg4 Qa8 23. Rad1 Kh7 24. Re3 Ra3 25. Rg3 Rg8 26. Nf1 Ra2 27. Re1 g6 28. Rh3 g5 29. Bc1 Rg6 30. Rhe3 Kg8 31. R3e2 Ra1 32. Ng3 Rb1 33. Rf1 Qa7 34. Nh5 Nc4 35. Rc2 Qa1 36. Qe2 Bd8 37. h3 Kh8 38. f4 c5 39. dxc5 gxf4 40. Nxf4 Rg7 41. Nd3 Kh7 42. Kh2 Bc7 43. g3 Bd8 44. Nf4 Bg5 45. c6 Nb6 46. Qd3+ Kg8 47. Nh5 Bxc1 48. Rg2 Kh8 49. Nxg7 Kxg7 50. c7 Qa7 51. Qxb1 Be3 52. Rxf7+ Kxf7 53. Ra2 Qxa2+ 54. Qxa2 Ke7 55. Qe2 d4 1-0Stockfish showed 1.13 at move 36, and after the deadly(?!) h-pawn push 37. h3 (with f4 to follow), jumped to 2.54, with Junior taking a few more moves to recognise the extent of the damage.

[BB+]

OK, here is what I produced for the Stockfish-Junior graph two posts previous. I ended up taking double logarithms as sign(y)*log(1+log(1+|y|)), and labelled the y-axis accordingly via sign(y)*(exp(exp(|y|)-1)-1). I also trimmed off the first 10 moves on the x-axis, and limited y in size to 10.0.

[Sarciness]

Shouldn't that be sin(y) ?

I think the idea of a logarithmic scale is an interesting one and I was also going to suggest starting the x-axis at 10 as there is no data before that point.

[BB+]

I'm not sure where a sine comes in. The y-axis is transformed vertically by twice applying the x->log(1+x) map (with care for absolute values).

Rybka just had a nice piece sac (move 43) against Sjeng:[White "Rybka 4.1"][Black "Sjeng"][Event "TCEC 3 Stage 2a"][Result "1-0"][Round "7.2"]1. d4 f5 2. g3 Nf6 3. Bg2 g6 4. Nf3 Bg7 5. O-O O-O 6. b4 Nc6 7. a3 d5 8. Bb2 Ne4 9. Nbd2 Be6 10. e3 a5 11. c3 Qd6 12. Qc2 Nd8 13. Rfc1 Nf7 14. bxa5 Rxa5 15. c4 c6 16. Ne1 Rfa8 17. Nd3 Bd7 18. Nb3 R5a7 19. f3 Nf6 20. Re1 dxc4 21. Qxc4 Nd5 22. e4 Nb6 23. Qc2 Bc8 24. e5 Qd8 25. f4 e6 26. Nbc5 Nd5 27. Nb4 Bf8 28. Rac1 Qa5 29. Qb3 Nd8 30. Red1 Nb6 31. Bf3 Na4 32. Nxa4 Qxa4 33. Qc3 Nf7 34. d5 cxd5 35. Rxd5 Qe8 36. Rd4 Bd7 37. Qc7 Bc6 38. Bxc6 bxc6 39. Qxc6 Qe7 40. Rc2 Nd8 41. Qf3 Qe8 42. Kg2 Nb7 43. Nd5 exd5 44. Qxd5+ Kh8 45. e6 Bg7 46. Rc7 Na5 47. Rxg7 Rxg7 48. Ra4 Kg8 49. Bxg7 Kxg7 50. Rxa5 Rxa5 51. Qxa5 Qc6+ 52. Kh3 Kf6 53. a4 Kxe6 54. Qb5 Qa8 55. a5 Kd6 56. Qb6+ Kd7 57. a6 Qf3 58. Qb7+ Qxb7 59. axb7 Kc7 60. Kh4 h5 61. Kg5 Kxb7 62. Kxg6 Kc7 1-0Black seemed to me to have the wrong strategic plan the whole game (or maybe, there was just no good plan to start), as playing on the queenside was playing into White's pawn structure. Rybka's central push with e4/e5 (moves 22-24) left Black with a position that is at best painful to defend long-term.

[ernest]

I'm not sure where a sine comes in.

[Sarciness]

Keep a linear scale, ditch eval for first 10 moves is what I think!

[BB+]

Just now, I realise there was a "sign" versus "sine" confusion - I did indeed meant the former, as for negative x, you first want to negate it. In realtime, the graphs are OK, but for those who don't match TCEC 24 hours a day, eyeballing a graph from a decisive game in the post-mortem can often not reveal too much.

Junior had another nice win against Sjeng, after White was preferred as much as 1.00 by Sjeng (move 26).
[White "Sjeng"][Black "Junior 12.5"][Round "8.3"][Event "TCEC 3 Stage 2a"][Result "0-1"]
1. d4 Nf6 2. Nf3 e6 3. Bf4 c5 4. e3 cxd4 5. exd4 b6 6. Nbd2 Be7 7. h3 O-O 8. Bd3 Ba6 9. Bxa6 Nxa6 10. O-O b5 11. c3 Nc7 12. a4 Ncd5 13. Bg3 Nh5 14. Bh2 a6 15. Qb3 Qb6 16. Ra2 h6 17. Rfa1 b4 18. c4 Ndf4 19. Qe3 g5 20. h4 f5 21. Bxf4 Nxf4 22. g3 Nh3+ 23. Kh2 g4 24. Ne5 Qb7 25. Qxh6 Bf6 26. Rf1 Bxe5 27. dxe5 Rf7 28. Raa1 Raf8 29. Rae1 f4 30. Qg6+ Rg7 31. Qd3 Ng5 32. Qd4 Rf5 33. Rg1 Rh7 34. Rd1 a5 35. b3 Qa8 36. Rg2 Kh8 37. c5 Qg8 38. Rgg1 Rhf7 39. c6 dxc6 40. Rg2 Qh7 41. Qd8+ Rf8 42. Qd3 Qh6 43. Qd4 R5f7 44. Qd3 c5 45. Kh1 Rf5 46. Re1 Nh3 47. Ne4 Rxe5 48. Qb1 Qg7 49. Qd1 Rd5 50. Qb1 Qg6 51. Kh2 Qf5 52. Qc1 Rd4 53. Nxc5 e5 54. gxf4 Nxf4 55. Qe3 Nxg2 56. Kxg2 Rf6 57. Kg1 Qc2 58. Rc1 Rd1+ 59. Rxd1 Qxd1+ 60. Kh2 Qf3 61. Kg1 Qxe3 62. fxe3 Rf3 63. Nb7 Rxe3 64. Nxa5 e4 65. Kh2 Rc3 66. Nc4 Rxc4 67. bxc4 b3 68. a5 b2 69. a6 0-1
[White "Rybka 4.1"][Black "Junior 12.5"][Event "TCEC 3 Stage 2a"][Round "9.2"][Result "1-0"]1. d4 Nf6 2. c4 e6 3. Nf3 c5 4. d5 d6 5. Nc3 exd5 6. cxd5 g6 7. g3 Bg7 8. Bg2 O-O 9. O-O Re8 10. Bf4 Na6 11. Nd2 Nh5 12. Be3 Nc7 13. a4 b6 14. Nc4 Ba6 15. Qb3 f5 16. Bf3 Qd7 17. Rfd1 Nf6 18. Bf4 Bxc4 19. Qxc4 a6 20. Qb3 b5 21. h4 b4 22. Nb1 Ne4 23. Nd2 Be5 24. e3 Nxd2 25. Rxd2 Bxf4 26. exf4 h5 27. Be2 Qg7 28. a5 Qd7 29. Bc4 Nb5 30. Rd3 Nd4 31. Rxd4 cxd4 32. Qxb4 Kh7 33. Qb6 Qc8 34. Qc6 Ra7 35. b3 Qb8 36. Qa4 Qc8 37. Kg2 Ree7 38. Rd1 Qc5 39. Qa1 Qb4 40. Kf1 Qc3 41. Kg2 Re4 42. Rc1 Qd2 43. Rd1 Qb4 44. Bd3 Ree7 45. Bc4 Qc3 46. Rc1 Qxa1 47. Rxa1 Re4 48. Rb1 Kg7 49. Rc1 Rc7 50. Ra1 Rxc4 51. bxc4 d3 52. Rd1 Rxc4 53. Rxd3 Rc5 54. Rb3 Rxd5 55. Rb7+ Kh6 56. Rb6 Rxa5 57. Rxd6 Ra3 58. Rb6 Kh7 59. Kh3 a5 60. Rb7+ Kh8 61. Rb6 Kh7 62. Rb7+ Kh8 63. Ra7 a4 64. Ra6 Kh7 65. Ra7+ Kg8 66. Ra6 Kf7 67. Ra7+ Ke6 68. Ra6+ Kd5 69. Rxg6 Kc4 70. Rc6+ Kb5 71. Rh6 Rc3 72. Rxh5 Rc5 73. Rh7 Kb4 74. h5 a3 75. Re7 a2 76. Re1 Rc2 77. f3 Rc3 78. Kh4 Rxf3 79. h6 Re3 80. Ra1 Ka3 81. h7 Re8 82. Kg5 Rf8 83. Kg6 Rd8 84. Kxf5 *
But then, just as I came in, Rybka had jumped fairly rapidly from 0.00 to 2.96 in a rook ending against Junior (which was almost drawn previously due to lots of near-0.00 scores and shuffling). The first move where Rybka had a non-0.00 score was 72. Rxh5, after which Junior originally thought Black was better. The scores were both 0.00 from move 50 to 63, but then Junior favoured Black a bit after the a4 push, but moving the King up so as to support the a-pawn backfired in the end.