Why no thinker talk?

[genorb]

The once that I have seen are at high depth. However, I did get the feeling that Thinker was doing something wrong as far as winboard protocol goes.

Is Thinker also an UCI engine? I tried the UCI protocole in Arena and Thinker simply doesn't work. I tried to install Thinker as an UCI engine in the Shredder Classic GUI and Shredder wrote that this is not an UCI engine...

[Tony Thomas]

The once that I have seen are at high depth. However, I did get the feeling that Thinker was doing something wrong as far as winboard protocol goes.

Is Thinker also an UCI engine? I tried the UCI protocole in Arena and Thinker simply doesn't work. I tried to install Thinker as an UCI engine in the Shredder Classic GUI and Shredder wrote that this is not an UCI engine...

Thinker only supports WB protocol. So as far as I know it doesnt support UCI. Lance said that he rewrote the xboard shell, as we noticed he did have the eternal pondering problem in the previous version. So I guess this is another bug he has to tackle.

[genorb]

Thanks Tony for explaining to me!

[Marc MP]

Hi Tony,

The passive (? It doesn't look passive to me... more like extremely aggressive!) Thinker is indeed a beast... Just to compare (on a Pentium 3.2 Ghz, 2m+2s, nunn matches):

Code: Select all

Rank Name               Elo    +    - games score oppo. draws 
   1 Thinker 5.1C      2800   25   25   300   55%  2767   29% 
   2 Fruit 2.2         2796   43   43   100   50%  2800   31% 
   3 Chess Tiger 2007  2766   43   44   100   45%  2800   28% 
   4 Spike 1.2         2737   43   44   100   41%  2800   28% 

   1 Thinker 5.1C      2800 300.0 (164.5 : 135.5)
                            100.0 ( 50.5 :  49.5) Fruit 2.2         2796
                            100.0 ( 55.0 :  45.0) Chess Tiger 2007  2766
                            100.0 ( 59.0 :  41.0) Spike 1.2         2737
   2 Fruit 2.2         2796 100.0 ( 49.5 :  50.5)
                            100.0 ( 49.5 :  50.5) Thinker 5.1C      2800
   3 Chess Tiger 2007  2766 100.0 ( 45.0 :  55.0)
                            100.0 ( 45.0 :  55.0) Thinker 5.1C      2800
   4 Spike 1.2         2737 100.0 ( 41.0 :  59.0)
                            100.0 ( 41.0 :  59.0) Thinker 5.1C      2800

A typical Thinker-"Passive" game:

[Event "Thinker Passive Gauntlet"]
[Site "Home"]
[Date "2008.3.10"]
[Round "1.1"]
[White "Thinker 5.1C"]
[Black "Fruit 2.2"]
[Result "1-0"]

1. e4 d5 2. exd5 Qxd5 3. Nc3 Qa5 4. d4 Nf6 5. Nf3 c6 6. Bc4 Bf5 7. O-O e6 {-0.94/11} 8. Bd2
Qb6 {-0.73/11} 9. Qe2 Bd6 {-0.35/10} 10. a4 a5 {-0.29/10} 11. Nh4 Bxc2 {-0.20/11}
12. Bg5 Qxb2 {-0.38/10} 13. Rac1 Qxc3 {-1.33/11} 14. Bxf6 gxf6 {-1.07/11} 15. Nf5
Bf4 {-0.95/11} 16. Rxc2 Qa3 {-1.47/11} 17. Bxe6 O-O {-2.36/12} 18. Bc4 Kh8
{-2.87/12} 19. Ra2 Qb4 {-4.71/13} 20. Rb2 Qa3 {-4.99/12} 21. Rxb7 Na6 {-5.29/11}
22. Bxf7 Rab8 {-5.41/10} 23. Qxa6 Rxb7 {-5.97/12} 24. Qxb7 Qb4 {-6.04/13} 25. Qd7
Qb8 {-6.32/12} 26. Qe6 Qd8 {-8.12/14} 27. Rb1 Bg5 {-8.26/13} 28. h4 Bxh4 {-8.26/11}
29. Nxh4 Qxd4 {-16.06/11} 30. Nf5 Qe5 {-99.84/10} 31. Qd7 Rb8 {-99.86/10} 32. Rd1
Qxf5 {-99.90/19} 33. Qxf5 Kg7 {-99.92/45} 34. Bg8 Kxg8 {-99.94/44} 35. Rd7 Re8
{-99.96/43} 36. Qxh7+ Kf8 {-99.98/4} 37. Qh8# {White wins} 1-0

[Uri Blass]

Hi,
the 1st match was not bad:

Code: Select all

1   (Passive) Thinker 64-bit  +17/-14/=19 53.00%   26.5/50
2   Glaurung 2.01 w32 1CPU    +14/-17/=19 47.00%   23.5/50

Difference to (active) Thinker is small, around +35 in this match.
Next opponent Frenzee Feb 08

I did say that you cant expect a 100 point improvement after seeing that you saw a 150 or so improvement with the active version.

100 elo is still possible and there are not enough games
I see that thinker does relatively better against frenzee and it is leading
18.5-11.5(frenzee is 30 elo better than default thinker)

Uri

[Werner]

Hi Uri,
I think till Sunday I will have around 300 games and we can see the differences. I have included Spike 1.2 Turin and Ktulu 8.0 in the match and here you can see the barrier (2770 ís too strong I think)

[Tony Thomas]

Hi,
the 1st match was not bad:

Code: Select all

1   (Passive) Thinker 64-bit  +17/-14/=19 53.00%   26.5/50
2   Glaurung 2.01 w32 1CPU    +14/-17/=19 47.00%   23.5/50

Difference to (active) Thinker is small, around +35 in this match.
Next opponent Frenzee Feb 08

I did say that you cant expect a 100 point improvement after seeing that you saw a 150 or so improvement with the active version.

100 elo is still possible and there are not enough games
I see that thinker does relatively better against frenzee and it is leading
18.5-11.5(frenzee is 30 elo better than default thinker)

Uri

If you didnt know, Thinker passive is sort of the default.

[Tony Thomas]

Hi Tony,

The passive (? It doesn't look passive to me... more like extremely aggressive!) Thinker is indeed a beast... Just to compare (on a Pentium 3.2 Ghz, 2m+2s, nunn matches):

Code: Select all

Rank Name               Elo    +    - games score oppo. draws 
   1 Thinker 5.1C      2800   25   25   300   55%  2767   29% 
   2 Fruit 2.2         2796   43   43   100   50%  2800   31% 
   3 Chess Tiger 2007  2766   43   44   100   45%  2800   28% 
   4 Spike 1.2         2737   43   44   100   41%  2800   28% 

   1 Thinker 5.1C      2800 300.0 (164.5 : 135.5)
                            100.0 ( 50.5 :  49.5) Fruit 2.2         2796
                            100.0 ( 55.0 :  45.0) Chess Tiger 2007  2766
                            100.0 ( 59.0 :  41.0) Spike 1.2         2737
   2 Fruit 2.2         2796 100.0 ( 49.5 :  50.5)
                            100.0 ( 49.5 :  50.5) Thinker 5.1C      2800
   3 Chess Tiger 2007  2766 100.0 ( 45.0 :  55.0)
                            100.0 ( 45.0 :  55.0) Thinker 5.1C      2800
   4 Spike 1.2         2737 100.0 ( 41.0 :  59.0)
                            100.0 ( 41.0 :  59.0) Thinker 5.1C      2800

A typical Thinker-"Passive" game:

[Event "Thinker Passive Gauntlet"]
[Site "Home"]
[Date "2008.3.10"]
[Round "1.1"]
[White "Thinker 5.1C"]
[Black "Fruit 2.2"]
[Result "1-0"]

1. e4 d5 2. exd5 Qxd5 3. Nc3 Qa5 4. d4 Nf6 5. Nf3 c6 6. Bc4 Bf5 7. O-O e6 {-0.94/11} 8. Bd2
Qb6 {-0.73/11} 9. Qe2 Bd6 {-0.35/10} 10. a4 a5 {-0.29/10} 11. Nh4 Bxc2 {-0.20/11}
12. Bg5 Qxb2 {-0.38/10} 13. Rac1 Qxc3 {-1.33/11} 14. Bxf6 gxf6 {-1.07/11} 15. Nf5
Bf4 {-0.95/11} 16. Rxc2 Qa3 {-1.47/11} 17. Bxe6 O-O {-2.36/12} 18. Bc4 Kh8
{-2.87/12} 19. Ra2 Qb4 {-4.71/13} 20. Rb2 Qa3 {-4.99/12} 21. Rxb7 Na6 {-5.29/11}
22. Bxf7 Rab8 {-5.41/10} 23. Qxa6 Rxb7 {-5.97/12} 24. Qxb7 Qb4 {-6.04/13} 25. Qd7
Qb8 {-6.32/12} 26. Qe6 Qd8 {-8.12/14} 27. Rb1 Bg5 {-8.26/13} 28. h4 Bxh4 {-8.26/11}
29. Nxh4 Qxd4 {-16.06/11} 30. Nf5 Qe5 {-99.84/10} 31. Qd7 Rb8 {-99.86/10} 32. Rd1
Qxf5 {-99.90/19} 33. Qxf5 Kg7 {-99.92/45} 34. Bg8 Kxg8 {-99.94/44} 35. Rd7 Re8
{-99.96/43} 36. Qxh7+ Kf8 {-99.98/4} 37. Qh8# {White wins} 1-0

I think the word passive is relative. It is passive only when compared to the active version.

[Werner]

If you didnt know, Thinker passive is sort of the default.

hm, I think I missed that - but at the moment it really looks like that:

My interim results are quite good for passive Thinker:
49% against Spike 1.2 Turin after 38 games and
52% against Ktulu 8.0 after 28 games

[Tony Thomas]

If you didnt know, Thinker passive is sort of the default.

hm, I think I missed that - but at the moment it really looks like that:

My interim results are quite good for passive Thinker:
49% against Spike 1.2 Turin after 38 games and
52% against Ktulu 8.0 after 28 games

Lance said that in a post here at Swami's thread. He said that Active uses algorithms mainly aimed towards humans were as passive performs better towards other engines. I would have expected better than 50% against both of the above mentioned engines, but we shall see.