Amyan 1.71 / some blitz results

[Mike S.]

3m+1s, D945/1*3,4 GHz (~P4)
128 MB hash each, ponder off
generic IM/GM book (*)
Arena 2.0.1, Vista 32 Bit
3/4- and R vs. R 5-piece tbs.
(for Rybka and Spike only)

*) book infos + downloadlink:
http://rybkaforum.net/cgi-bin/rybkaforu ... pid=164933

Code: Select all

Amyan 1.71 WB - Pro Deo 1.6   4.0 - 8.0    +3/-7/=2    33.33%
Amyan 1.71 WB - Rybka 2.3.2a  1.0 - 11.0   +1/-11/=0    8.33%
Amyan 1.71 WB - Spike 1.2     2.5 - 9.5    +1/-8/=3    20.83%

If my calculations using CCRL 40/4 ratings are correct, than the expectation for the old version 1.597 would be ~15% or 5.5/36. 1.71 scored 7.5 which is a good sign, although not statistically reliable yet.

On my system with Vista 32 Bit and in Arena 2.0.1, if I play a normal game user vs. engine, WinBoard-Amyan 1.71 cannot ponder. If configured as UCI engine, ponder works. (The tests were ponder off anyway, so I used it as WB. engine.)

Some examples. Big skalp :

[Event "Amyan171-Testmatch"]
[Site "D945 3.4 GHz (~dual P4)"]
[Date "2009.10.17"]
[Round "5"]
[White "Amyan 1.71 WB"]
[Black "Rybka 2.3.2a"]
[Result "1-0"]
[ECO "A36"]
[TimeControl "180+1"]
[Termination "normal"]
[PlyCount "108"]

1. e4 c5 2. Nc3 Nc6 3. d3 g6 4. g3 Bg7 5. Nge2 d6 6. Bg2 e5
7. O-O Nge7 8. f4 O-O 9. fxe5 Bxe5 10. Bh6 Re8 11. Qd2 Qa5
12. Rf2 Nd4 13. Nf4 Bg4 14. Raf1 Rab8 15. a3 Qb6 16. b4
cxb4 17. Ncd5 Ne2+ 18. Kh1 Bc3 19. Nxc3 bxc3 20. Qe1 Nxf4
21. Bxf4 Rbc8 22. Be3 Qb2 23. Bc1 Qb5 24. Rxf7 Nf5 25. Rf6
Qe5 26. Bg5 Ng7 27. Qe3 Be6 28. R6f2 Nh5 29. Bh6 Qa5 30. g4
Ng7 31. Qd4 Qe5 32. Qxa7 Ra8 33. Qxb7 Reb8 34. Qc6 Rc8
35. Qb6 Rcb8 36. Qe3 Bxg4 37. d4 Qe7 38. Qxc3 Rc8 39. Qe3
Nh5 40. e5 Rab8 41. Bd5+ Be6 42. Qf3 Kh8 43. Bxe6 Qxe6
44. Qf8+ Rxf8 45. Rxf8+ Qg8 46. Rxg8+ Kxg8 47. exd6 Rd8
48. c4 Ng7 49. Bxg7 Kxg7 50. c5 g5 51. d5 Kg8 52. c6 Kg7
53. c7 Rc8 54. d7 Rg8 1-0

Oops:

[Event "Amyan171-Testmatch"]
[Site "D945 3.4 GHz (~dual P4)"]
[Date "2009.10.17"]
[Round "1"]
[White "Amyan 1.71 WB"]
[Black "Pro Deo 1.6"]
[Result "0-1"]
[ECO "D07"]
[TimeControl "180+1"]
[Termination "normal"]
[PlyCount "24"]

1. d4 d5 2. c4 Bf5 3. cxd5 Bxb1 {-0.20/11 6} 4. Rxb1 Qxd5 {-0.16/11 6} 5.
e3 e5 {-0.04/11 7} 6. Qf3 {+0.32/10 3} Bb4+ {+0.91/10 4} 7. Kd1 {0.00/12 2}
e4 {+0.57/10 7} 8. Qg4 {+0.11/12 3} Nc6 {+0.16/10 6} 9. Qxg7 {+0.40/11 2}
Qxa2 {+3.72/10 4} 10. Kc2 {-1.50/12 3} Bf8 {+3.93/11 7} 11. Qxh8 {-1.88/13
2} Nb4+ {+6.30/10 9} 12. Kc3 {-3.87/13 7} Qxb1 {+6.60/5 0} 13. ... {White
resigns} 0-1

Pruning oops(?) in Pro Deo - Amyan, after 17...h5 (-0.26/11):
[d]r1b2rk1/1p3pp1/p1p1p3/3pP2p/1bPP2N1/3B1N1P/1P3P2/2QK4 w - - 0 18
White has 18.Nf6+!! #5

The configuration in style.ini was:

Code: Select all

HASHMEM            =128 
BOOK               =0
RENDITION          =1
RENDITION MARGIN   =7
FIXED SEARCH DEPTH =0
OPPONNENT TYPE     =1

(rest defaults)

[adieguez]

hi, yes, prunning oops, not much serious but it is actually a bug because I wrote something not exactly like I pretended. The impact of the bug should be really really small, am not saying it makes amyan play weaker, I don't know, I just wrote something not as I pretended, so I will not post a bug-fix just for this.