The occasional victory

[sje]

It's amazing that the current version of the program can win any games at all considering it has no positional evaluation terms for scoring kings or relationships between a king and other men. Also, the pawn evaluation is mostly limited to giving bonuses for advancement and for passed status. There's no concept of game stage, nor of material balance beyond a simple sum of man values.

However, the opening book works quite well, and the tablebases are probed at every node where there are five or fewer men. This probing is fast because the files are stored in uncompressed format on a solid state drive, and the probe results have their own transposition table.

----

Within a week or so I'll have all of the XBoard protocol two features working and well tested except for some ICS specific items which require access to an ICS. I'm hoping I can use my old accounts on HGM's server for this -- and if I can remember the IDs and passwords.

After I'm confident that all XBoard related operations are working smoothly and without need of manual supervision, I'll start adding more features to Symbolic: pondering, decent move ordering, a comprehensive evaluation function. etc. At that point, I'll get the program running once again on FICS and ICC so that anyone within 800 elo can take a shot at it.

[cdani]

It's amazing that the current version of the program can win any games at all considering it has no positional evaluation terms for scoring kings or relationships between a king and other men. ...

The new version I just started of Andscacs, having maybe 2100 and knowing only material+mobility, won a game (losing +700) against the full version (+3000). If you follow the game you see some pawn sacrifices of the good one that weakens a lot the position of the weak one, but somehow nothing happens and also in the end a bishop is half trapped in a way that the two bishops are distracted enough to decide the game. Just an statistically possible chain of bad luck and some misevaluations that happens only in one in a lot of games... And of course the inherent lack of real understanding of engines.

[sje]

And the occasional draw as well.

I'm still adding a few more items to Symbolic's XBoard interface, lke handling playother and picking up the opponent's entire name instead of just its first token.
[pgn][Event "XBoard Event"]
[Site "gail"]
[Date "2015-08-28"]
[Round "1"]
[White "Symbolic 2015-08-28"]
[Black "Sjeng 11.2"]
[Result "1/2-1/2"]
[Termination "White draws by position repetition"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 d6 4. Nf3 Nxe4 5. Qe2 Qe7 6. d3 Nf6 7. Qxe7+ Bxe7 8. Nc3 O-O 9. Be3 Nc6
10. Kd2 Bf5 11. Be2 Rfb8 12. a4 a5 13. Nd4 Nxd4 14. Bxd4 c5 15. Be3 d5 16. d4 cxd4 17. Bxd4 Ne4+
18. Kd1 Nxc3+ 19. Bxc3 Bc5 20. g4 Bg6 21. f4 Be4 22. Re1 Rc8 23. Rc1 Bf2 24. Rf1 Ba7 25. Bf3 Rc4
26. Ra1 d4 27. Bxa5 Bxc2+ 28. Kd2 Bb8 29. Bd5 Rxa4 30. Rae1 Bd6 31. Bxb7 Rb8 32. Kxc2 Rxa5 33. Be4
Ra2 34. Rb1 Ba3 35. h4 Rb5 36. Bc6 Raxb2+ 37. Rxb2 Rxb2+ 38. Kd3 Bc5 39. Bd5 Kf8 40. Rc1 Rb5 41.
Kc4 Ba3 42. Ra1 Rb4+ 43. Kd3 Ra4 44. Ra2 Ke7 45. Bb3 Ra5 46. g5 f5 47. Bg8 h5 48. Kc4 Kf8 49. Be6
Ke7 50. Bg8 Kf8 51. Be6 Ke7 52. Bg8 1/2-1/2[/pgn]
Note: the above PGN game and the others posted all come from Symbolic's PGN log, not from XBoard's output.

It would be nice if Sjeng didn't pepper my home directory with its learning data files.

[sje]

Symbolic has the Capablanca property; Capablanca was never checkmated in serious play, and Symbolic has never been checkmated playing on an ICS.

When it comes time to play the result of a move selection, Symbolic runs a quick test to see if after the move were to be played if a mate-in-1 or a mate-in-2 is available to the opponent. If so, the program skips playing its move and graciously resigns like a true gentleman.

Symbolic also has the EagleEye property; it will never miss playing a mate-in-3 (or faster) if one is available. I wonder which human players can state the same; most GMs and IMs, I suppose.

Symbolic's EagleEye runs prior to a full search; if a mating move is present from EagleEye or the tablebases, then the main search is skipped entirely.

[sje]

Example resignation

Congratulations to tscp:
[pgn][Event "XBoard Event"]
[Site "gail"]
[Date "2015-08-29"]
[Round "14"]
[White "tscp"]
[Black "Symbolic 2015-08-29"]
[Result "1-0"]
[Termination "Black resigns"]

1. d4 Nf6 2. Nf3 g6 3. c4 Bg7 4. g3 O-O 5. Bg2 d6 6. O-O Nc6 7. Nc3 Bf5 8. d5 Na5 9. Qa4 c5 10. Bd2
b6 11. Nd1 Bd7 12. Qc2 Bf5 13. Qc1 Nb7 14. Ng5 h6 15. e4 Bd7 16. Nh3 Nxe4 17. Bxh6 Nf6 18. Bxg7
Kxg7 19. Nf4 Rh8 20. Re1 Na5 21. Qc2 a6 22. Ne3 b5 23. b3 g5 24. Nd3 Qb6 25. Qc1 Rh5 26. g4 Nxg4
27. Bf3 bxc4 28. bxc4 f5 29. Nxg4 fxg4 30. Rxe7+ Kg8 31. Rxd7 Ra7 32. Rxa7 Qxa7 33. Bxg4 Rh7 34.
Qxg5+ Rg7 35. Qf4 Qf7 36. Qg3 Qg6 37. h3 Nxc4 38. Rb1 Kh7 39. f4 Qf6 40. Qf2 a5 41. f5 Re7 42. Nxc5
dxc5 43. Qxc5 Nd6 44. a4 Ne4 45. Qb6 Rg7 46. Qxf6 Nxf6 47. Kf2 Nxg4+ 48. hxg4 Rxg4 49. Ke3 Rxa4 50.
d6 Ra3+ 51. Ke4 Ra2 52. d7 Rd2 53. Rb7 Kh6 54. f6 Rd1 55. f7 Kg7 56. Rb8 Re1+ 57. Kf3 Kg6 58. f8=Q
Rf1+ 59. Ke4 Re1+ 60. Kd3 Rd1+ 61. Ke3 Re1+ 62. Kd2 Re7 63. Qxe7 Kf5 64. d8=Q 1-0[/pgn]