Twisted Logic against CB oldies

[Mike S.]

8m+2s, AthlonXP@1439 MHz
128 MB hash each (Tiger 96 MB)
+8 MB pawn hash for TL
Balanced-16.CTG (normal, min.3)
ponder off, Fritz 8-GUI
3- and 4-piece tablebases
(not accessed by F 5.32 and TL)

Code: Select all

Twisted Logic 20080620p8   - Fritz 5.32                4.5 - 1.5    +3/-0/=3    75.00%
Twisted Logic 20080620p8   - Fritz f.Fun 4 (8.130)     0.0 - 6.0    +0/-6/=0    0.00%
Twisted Logic 20080620p8   - Gambit Tiger 2.0          2.0 - 4.0    +1/-3/=2    33.33%
Twisted Logic 20080620p8   - Nimzo 8                   3.0 - 3.0    +1/-1/=4    50.00%
Twisted Logic 20080620p8   - Shredder 7.SE             1.5 - 4.5    +0/-3/=3    25.00%

p8 = 8 MB pawn hash (19)

TL won the mini match against Fritz 5.32 4,5-1,5, but scored only two wins (and 9 draws) in the remaining 24 games, against these engines. After a comparison with CCRL and CEGT 40/4m ratings (incl. some estimations), I think this was a blitz performance of ~2600(CCRL) or ~2570(CEGT). This is a level where quite good engines are, for example Yace, Little Goliath or Anaconda. (Although, I guess these three would rank much higher hadn't their developement stopped years ago.)

Technically, everything ran perfectly. TL usually diplayed much bigger depths than the opponents. I post two examples in the follow-up posting.

[Mike S.]

[Event "TL 20. Juni"]
[Site "?"]
[Date "2008.06.21"]
[Round "1.1"]
[White "Twisted Logic 20080620p8"]
[Black "Fritz 5.32"]
[Result "1-0"]
[ECO "E69"]
[PlyCount "75"]
[TimeControl "480+2"]

1. c4 {B/0 0} e5 {B/0 1} 2. g3 {B/0 0} g6 {B/0 1} 3. d4 {B/0 0} d6 {B/0 0} 4.
Nc3 {0.67/16 51} Nd7 {(Le7) 0.12/11 21} 5. Nf3 {0.78/13 16} Bg7 {
(Sgf6) 0.12/12 52} 6. Bg2 {B/0 0} Ngf6 {B/0 0} 7. O-O {B/0 0} O-O {B/0 0} 8. e4
{B/0 0} c6 {B/0 0} 9. b3 {B/0 0} Re8 {B/0 0} 10. h3 {B/0 0} exd4 {0.00/11 20}
11. Nxd4 {B/0 0} Nc5 {B/0 0} 12. Re1 {B/0 0} Ne6 {0.06/10 16} 13. Be3 {
0.36/16 46} Nxd4 {0.06/10 12} 14. Bxd4 {0.32/16 14} Be6 {(c5) 0.22/11 38} 15.
Ne2 {(Df3) 0.54/14 14} c5 {(Ld7) 0.22/10 14} 16. Bc3 {(Lb2) 0.66/14 13} Qd7 {
(a5) 0.22/11 11} 17. g4 {(e5) 0.90/14 13} h5 {(De7) 0.34/11 29} 18. g5 {
(gxh5) 0.99/15 13} Nh7 {0.37/12 11} 19. Bxg7 {1.05/15 12} Kxg7 {0.09/10 3} 20.
Qd2 {1.04/14 12} Rad8 {(De7) 0.28/11 17} 21. h4 {1.02/13 23} f6 {0.31/12 22}
22. Nf4 {0.89/13 11} a6 {(Lf7) 0.12/10 15} 23. Qc3 {(Tad1) 1.04/12 11} Bf7 {
(Df7) 0.25/10 10} 24. Bh3 {0.58/15 26} Qc7 {(De7) 0.25/11 16} 25. Nd5 {
(Te3) 1.45/14 10} Bxd5 {0.31/12 26} 26. exd5 {1.30/15 9} b6 {(b5) 0.41/11 14}
27. Be6 {1.76/13 9} b5 {0.37/11 12} 28. f4 {(a3) 1.79/14 22} b4 {
(bxc4) 0.62/11 13} 29. Qb2 {(gxf6+) 1.80/14 8} Qa5 {(De7) 0.97/11 8} 30. f5 {
(Te2) 3.81/15 8} Qa3 {1.84/11 7} 31. Qf2 {4.34/15 8} Rf8 {(fxg5) 2.53/12 19}
32. fxg6 {4.69/16 8} fxg5 {2.84/11 12} 33. Bf5 {5.02/16 8} Rd7 {3.62/12 16} 34.
Qc2 {(gxh7) 5.68/16 7} Rc7 {(Tb7) 4.00/11 6} 35. gxh7 {5.70/14 7} Rcf7 {
(gxh4) 4.12/10 5} 36. Bg6 {6.14/15 7} Rf4 {(Tf3) 5.12/11 6} 37. Bxh5 {8.46/14 7
} Kh8 {6.06/11 8} 38. hxg5 {10.05/16 7} 1-0

[Event "TL 20. Juni"]
[Site "?"]
[Date "2008.06.21"]
[Round "6.2"]
[White "Twisted Logic 20080620p8"]
[Black "Fritz for Fun 4"]
[Result "0-1"]
[ECO "B01"]
[PlyCount "54"]
[TimeControl "480+2"]

1. e4 {B/0 0} d5 {B/0 0} 2. exd5 {B/0 0} Qxd5 {B/0 0} 3. Nf3 {B/0 0} Bg4 {
0.05/12 20} 4. Be2 {B/0 0} Nc6 {-0.14/12 16} 5. d4 {B/0 0} O-O-O {-0.23/11 32}
6. Be3 {B/0 0} e5 {B/0 0} 7. c4 {B/0 0} Qa5+ {B/0 0} 8. Bd2 {B/0 0} Bb4 {B/0 0}
9. d5 {B/0 0} Bxf3 {B/0 0} 10. Bxf3 {B/0 0} Nd4 {-0.01/12 26} 11. O-O {
(Lc3) -0.09/15 53} Nxf3+ {(Sf6) -0.95/11 18} 12. gxf3 {0.17/15 16} Bxd2 {
(Sf6) -0.92/11 24} 13. Nxd2 {(Dxd2) 0.11/15 15} Nh6 {(Sf6) -0.93/12 18} 14. Qc2
{(Te1) 0.17/14 23} Rhe8 {(g6) -0.86/11 11} 15. a3 {(Tfe1) 0.11/14 14} Qb6 {
(g6) -1.05/12 24} 16. Rad1 {(Kh1) 0.03/15 21} Qf6 {(Dg6+) -1.13/12 16} 17. Qxh7
{(Tfe1) 0.05/15 13} Qg5+ {(Th8) -1.58/11 6} 18. Kh1 {0.90/8 0} g6 {
(Df4) -2.13/12 8} 19. Rfe1 {-0.38/15 13} Qh5 {(Dh4) -3.06/11 4} 20. Qg7 {
(Kg2) -1.39/19 135} Nf5 {-2.65/11 4} 21. Qf6 {-1.71/18 8} Rd6 {-4.78/11 4} 22.
Qxf7 {(Txe5) -1.71/17 8} Rh8 {-5.64/11 3} 23. Nf1 {-5.03/18 80} Qh3 {-6.85/11 5
} 24. Rxe5 {-5.47/19 50} Qxf3+ {-6.38/11 3} 25. Kg1 {-4.32/8 0} Qg4+ {
-6.87/11 3} 26. Kh1 {-5.62/17 4} Qxd1 {-6.87/11 5} 27. Re8+ {-5.93/16 4} Rxe8 {
-7.00/12 3} 0-1

[Edsel Apostol]

Hi Mike,

It seems like a poor performance by Twisted Logic.

Would you try again with the default setting only? Maybe the engine doesn't like having a big pawn hash, or there's a bug in its implementation.

[Dr.Wael Deeb]

Hi Mike,

It seems like a poor performance by Twisted Logic.

Would you try again with the default setting only? Maybe the engine doesn't like having a big pawn hash, or there's a bug in its implementation.

I agree,it should have done much better....
Maybe an adapter issue if used

[Mike S.]

Since TL is UCI compatible, I didn't need to use an external adapter in the Fritz(8) GUI, which is very stable and reliable. I also checked the CPU load and the engines got ~97% each (ponder off).

As mentioned, I didn't notice a technical problem because TL displayed huge search depths, ran "normally" and scored at least some wins and draws. But I will repeat the tournament today with default pawn hash size. The only other difference will be that different openings will be played (but the same book will be used).

I hope to have the results after ~10 hours from now.

[Dr.Wael Deeb]

Since TL is UCI compatible, I didn't need to use an external adapter in the Fritz(8) GUI, which is very stable and reliable. I also checked the CPU load and the engines got ~97% each (ponder off).

As mentioned, I didn't notice a technical problem because TL displayed huge search depths, ran "normally" and scored at least some wins and draws. But I will repeat the tournament today with default pawn hash size. The only other difference will be that different openings will be played (but the same book will be used).

I hope to have the results after ~10 hours from now.

Thanks Mike for the clarification,I know that you are an experienced tester,but still....these results could be better I assume

[Werner]

Hi Mike,

It seems like a poor performance by Twisted Logic.

Would you try again with the default setting only? Maybe the engine doesn't like having a big pawn hash, or there's a bug in its implementation.

Hi Edsel,
I have very good results till now for CEGT 40/20. Only the match against Smarthink Moscow looks not good (but I have only 8 games till now).

congrats!

[Edsel Apostol]

Hi Werner,

Thanks.

I'm looking forward to see the latest Twisted Logic version in the CEGT rating list.