Evaluation without material count

[Henk]

I wondered if it would be possible to play a reasonable game without counting material.
In the game below evaluation only computes a sloppy mobility term for each piece type and no other terms added,

Unfortunately I had to add a material term for the pawns. Otherwise it would give away too many pawns.
So I still have to fix that to play without counting any material.

O wait in the end game evaluation also contains two extra terms: one for passers and king position.


[pgn] [Event "Computer Chess Game"] [Site "HP"] [Date "2019.03.22"] [Round "-"] [White "SchaakMachine"] [Black "Fairy-Max 4.8S"] [Result "1-0"] [TimeControl "300"] [Annotator "1. +9.20 1... +0.14"] 1. d4 {+9.20/9} f5 {+0.14/8 8} 2. Nc3 {+13.40/9 6} d6 {+0.08/8 4} 3. e4 {+13.40/8 6} Nf6 {+0.17/7 4} 4. exf5 {+15.20/8 6} Bxf5 {+0.01/9 4} 5. Nf3 {+16.00/7 6} Nc6 {-0.10/8 4} 6. Bd3 {+13.00/7 5} Qd7 {-0.01/9 6} 7. Bg5 {+12.20/7 5} Nb4 {+0.33/9 5} 8. h4 {+13.20/8 5} Nxd3+ {+0.51/9 8} 9. cxd3 {+12.61/8 5} O-O-O {+0.55/9 6} 10. h5 {+15.60/8 5} h6 {+0.75/8 3} 11. Bc1 {+14.00/8 5} Bg4 {+0.81/9 13} 12. Rh4 {+14.40/8 5} Bxh5 {+1.03/8 4} 13. Be3 {+14.80/8 5} Qe8 {+1.23/8 7} 14. Qb3 {+13.82/8 5} Qg6 {+1.52/9 6} 15. Rxh5 {+17.42/8 5} Qxh5 {+2.63/8 2.9} 16. Kd2 {+16.16/8 4} Ng4 {+2.81/8 3} 17. Qe6+ {+20.28/8 4} Rd7 {+3.07/8 4} 18. b4 {+18.28/8 4} Nxe3 {+3.07/9 5} 19. Kxe3 {+17.60/8 4} c6 {+2.94/10 9} 20. b5 {+16.80/8 4} c5 {+2.90/10 7} 21. Ne4 {+19.69/8 4} b6 {+2.87/10 3} 22. g4 {+19.90/8 4} Qh3 {+2.94/10 2.4} 23. dxc5 {+21.31/8 4} bxc5 {+1.74/10 3} 24. Qd5 {+19.10/8 4} Rb7 {+2.79/9 2.3} 25. Rg1 {+18.29/8 4} Rb8 {+2.25/9 4} 26. Qf7 {+21.60/7 4} Kd8 {+2.48/10 4} 27. Nc3 {+19.09/8 4} g5 {+2.55/10 2.8} 28. Ke2 {+15.90/8 3} Rc8 {+2.41/10 3} 29. Ne4 {+17.10/8 3} Kc7 {+2.20/10 6} 30. Nf6 {+22.69/7 3} c4 {+1.80/8 2.2} 31. dxc4 {+24.40/7 3} Rd8 {+0.72/9 5} 32. Nd5+ {+26.50/7 3} Kb7 {+0.75/9 1.9} 33. Ne3 {+23.70/8 3} Kc8 {+0.99/8 2.6} 34. b6 {+27.10/7 3} Kb8 {-0.46/9 4} 35. Rb1 {+25.60/8 3} d5 {-1.77/10 7} 36. cxd5 {+29.30/8 3} Kb7 {-0.34/9 1.7} 37. bxa7+ {+30.80/8 3} Kxa7 {-9.29/10 2.3} 38. Nd4 {+30.00/8 2.9} Qh2 {-7.45/10 8} 39. Nc6+ {+35.30/8 2.8} Ka8 {-9.52/10 2.3} 40. Nxd8 {+33.29/7 2.8} Qd6 {-6.79/10 1.8} 41. Ne6 {+31.24/7 2.7} Qa6+ {-4.67/10 3} 42. Kf3 {+38.84/8 2.7} Bg7 {-7.94/10 3} 43. Qxg7 {+39.25/8 2.6} Rc8 {-9.83/11 1.7} 44. Qxe7 {+36.92/8 2.6} Qa7 {-13.52/11 2.3} 45. Qf6 {+37.24/7 2.5} Rb8 {-8.43/10 4} 46. Rc1 {+36.88/7 2.5} Qa5 {-8.12/9 1.4} 47. Nc7+ {+38.04/8 2.4} Ka7 {-11.89/10 1.6} 48. Qd4+ {+38.07/8 2.4} Rb6 {-15.87/11 2.0} 49. Ne6 {+44.11/7 2.4} Qb4 {-10.29/9 1.4} 50. Nc4 {+43.27/7 2.4} h5 {-11.00/9 1.7} 51. gxh5 {+44.94/7 2.2} Kb7 {-12.80/9 1.6} 52. Qxb6+ {+47.33/7 2.2} Qxb6 {-18.41/13 1.3} 53. Nxb6 {+67.12/9 2.2} Kxb6 {-79.93/17 1.4} 54. h6 {+7.45/8 2.1} g4+ {-79.94/14 1.5} 55. Kxg4 {+8.27/9 2.1} Kb5 {-79.95/14 6} 56. h7 {+8.35/8 2.0} Ka5 {-79.96/14 1.2} 57. h8=Q {+2621.35/8 2.0} Ka6 {-79.97/21 1.2} 58. Rb1 {+2621.37/7 1.9} Ka7 {-79.98/28 0.2} 59. d6 {+2621.39/7 1.9} Ka6 {-79.99/28 0.1} 60. Qa8# {+2621.41/7 1.9} {Xboard adjudication: Checkmate} 1-0 [/pgn]

[Henk]

Hmm appears to be a lucky game. For in next game I see it gives away too many pieces. Maybe I can change its only mobility term to fix that.
Not very promising.

[Henk]

Hi hi hi without any evaluation term:

[pgn] [Event "Computer Chess Game"] [Site "HP"] [Date "2019.03.22"] [Round "-"] [White "SchaakMachine"] [Black "Fairy-Max 4.8S"] [Result "1/2-1/2"] [TimeControl "300"] [Annotator "1. +0.00 1... +0.13"] 1. Na3 {+0.00/17} Nh6 {+0.13/9 9} 2. Rb1 {+0.00/17 6} c5 {+0.19/9 8} 3. Ra1 {+0.00/16 6} b6 {+0.07/9 9} 4. Rb1 {+0.00/16 6} Nc6 {-0.01/9 5} 5. Ra1 {+0.00/16 6} Nb8 {-0.01/10 5} 6. Rb1 {+0.00/20 5} Nc6 {+0.00/11 14} 7. Ra1 {+0.00/20 5} Nb8 {-0.01/11 7} {XBoard adjudication: repetition draw} 1/2-1/2 [/pgn]

This result is repeatable so a bug in Fairy max engine ?!

[pgn] [Event "Computer Chess Game"] [Site "HP"] [Date "2019.03.22"] [Round "-"] [White "SchaakMachine"] [Black "Fairy-Max 4.8S"] [Result "1/2-1/2"] [TimeControl "300"] [Annotator "1. +0.00 1... +0.17"] 1. Na3 {+0.00/17} Nf6 {+0.17/9 9} 2. Rb1 {+0.00/16 6} h5 {+0.19/9 9} 3. Ra1 {+0.00/17 6} Rh7 {+0.10/9 9} 4. Rb1 {+0.00/17 6} c5 {+0.06/9 4} 5. Ra1 {+0.00/16 6} Nc6 {-0.02/10 17} 6. Rb1 {+0.00/15 5} Nb4 {-0.01/9 5} 7. Ra1 {+0.00/15 5} Nc6 {+0.00/11 6} 8. Rb1 {+0.00/20 5} Nb4 {-0.01/10 4} 9. Ra1 {+0.00/20 5} Nc6 {+0.00/12 6} {XBoard adjudication: repetition draw} 1/2-1/2 [/pgn]

[hgm]

Well, the 'bug' would be that Fairy-Max doesn't have contempt, and quasi-randomizes its first few moves. The randomization often picks moves that Fairy-Max itself would consider bad, and especially with black this then brings it in a position where it considers itself at a disadvantage. It is then happy with a draw, and when the opponent reverts his previous move, it will also do that, to salvage a draw ffrom a poor position.