Rebel TNG (version 1.0) approximately 2950....

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
lucasart
Posts: 3232
Joined: Mon May 31, 2010 1:29 pm
Full name: lucasart

Re: Rebel TNG (version 1.0) approximately 2950....

Post by lucasart »

Rebel wrote:
Werner wrote:
Graham Banks wrote:Pretty sure that ProDeo 1.88 was okay too?
Graham.
Hi Graham, I am not sure - how to stop this engine using all the databases?

PS: here Rebel TNG uses 58% of a quad - more than 2 CPUs

regards
Werner
You can not turn off the database search, even if you would manage it will play exactly as ProDeo 1.88. If you don't like the program in its current form siimply don't test it :wink:
That's total cheating. You've basically hardcoded a massive database of precalc strong moves (ie. a large opening book).

to measure the real value of the engine, i recommend chess960. that should drastically reduce the db hit rate.
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Rebel »

lucasart wrote:That's total cheating.
Is this really you Lucas? :lol:

I have made a bet with my wife that the first accusation of cheating would come on the first day. I lost the bet :wink: but I did not expect it coming from an experienced programmer such as yourself.
You've basically hardcoded a massive database of precalc strong moves (ie. a large opening book).
Then where is this massive database?
to measure the real value of the engine, i recommend chess960. that should drastically reduce the db hit rate.

Checkout the root directory, the so called massive :roll: database is only about 2Mb, no single move inside, just bits and bits, you know those particles you can put on and off and more bits serving as indexes. Each pattern (currently) contains 128 bits and each bit (or a group of max 4 bits) is referring to particular chessknowledge. Then for each matching pattern a bonus is calculated for the evaluation function.

If you would had read my pages you would have known that.
User avatar
Werner
Posts: 2871
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Werner »

Hi,
in CEGT 40/20 we are testing the engines e.g. with 1 CPU, using a short opening book - and we are using tbs (4 or 5men). Tbs are nothing other than a large database. The engines use them under our conditions - with 1CPU. In earlier times I remember the kns dropped sometimes dramatically down when tbs are accesses.
So, why not say now Rebel is using a similar database (not only for endgames - earlier). If the engine is using 1CPU for all that - we have similar conditions.
I did following experement:
Shredder GUI uses only CPU 3 on my quad and so will do the engines inside this GUI. Then I started a tournment
Rebel TNG - ICE 1.0 x64 1CPU with tbs. The first game ran without Problems. In the taskmanager I saw Rebel.exe and pattern.exe using together the same CPU. Both running with 50% of the CPU when starting - and when pattern.exe is finished Rebel.exe use 100%.

here is the first game:
regards
Werner
[pgn]
[Event "40 Züge in 8 min"]
[Site "ProDeo2, I7"]
[Date "2014.04.01"]
[Round "1.1"]
[White "Rebel TNG"]
[Black "iCE 1.0 x64"]
[Result "1/2-1/2"]

1. Nf3 {Buch 0s} c5 {Buch 0s} 2. g3 {Buch 0s} Nc6 {Buch 0s}
3. Bg2 {Buch 0s} d5 {Buch 0s} 4. O-O {Buch 0s} e6 {Buch 0s}
5. d3 {Buch 0s} Bd6 {Buch 0s} 6. Nbd2 {Buch 0s} Nge7 {Buch
0s} 7. e4 {Buch 0s} O-O {Buch 0s} 8. Re1 {Buch 0s} Qb6
{0.00/14 13s} 9. exd5 {+0.50/15 14s} exd5 {+0.17/18 14s}
10. c4 {+0.46/15 14s} Bg4 {+0.23/18 14s} 11. h3 {+0.16/14
16s (cxd5)} Bh5 {-0.43/17 14s} 12. cxd5 {+0.13/14 17s} Nxd5
{-0.35/18 15s} 13. Nc4 {+0.47/15 13s} Qc7 {-0.35/16 14s}
14. Nxd6 {+0.14/14 13s (Lg5)} Qxd6 {-0.42/16 14s} 15. d4
{+0.38/15 15s (Db3)} Rad8 {-0.73/17 16s (Tae8)} 16. dxc5
{+0.22/15 22s} Qxc5 {-0.69/17 14s} 17. Bd2 {+0.14/15 12s}
Ndb4 {-0.46/18 29s} 18. g4 {+0.17/16 15s (Lxb4)} Bg6
{-0.66/17 13s} 19. Qb3 {+0.21/15 14s} Nd3 {-0.57/17 14s}
20. Be3 {+0.17/14 12s} Qb4 {-0.47/17 11s} 21. Red1
{+0.28/15 12s} Nxb2 {-0.58/18 13s} 22. Rxd8 {+0.25/17 14s
(Dxb4)} Rxd8 {-0.39/20 13s} 23. Qxb4 {+0.22/17 16s} Nxb4
{-0.33/19 13s} 24. Bxa7 {+0.56/16 11s} Be4 {+0.01/20 42s
(Sc4)} 25. Ne1 {+0.54/17 20s (Se5)} Bxg2 {-0.39/21 12s}
26. Kxg2 {+0.72/15 4s} Rd1 {-0.36/22 18s (Ta8)} 27. Rxd1
{+0.69/19 13s} Nxd1 {-0.30/22 11s} 28. a3 {+0.70/18 12s}
Nc6 {-0.16/22 10s (Sd5)} 29. Bc5 {+0.68/17 10s} f6
{-0.12/21 35s (Sc3)} 30. f4 {+0.72/17 11s (Sd3)} Kf7
{-0.12/19 9s} 31. f5 {+0.53/17 17s (Sd3)} Ne5 {-0.26/20 9s
(Sc3)} 32. Nf3 {+0.64/17 11s (Ld4)} Nc4 {-0.21/19 10s}
33. Nd4 {+0.62/18 16s} b6 {-0.16/20 9s (g6)} 34. Bb4
{+0.67/19 17s} Ne5 {0.00/21 9s (g6)} 35. Kg3 {+0.73/17 10s}
g6 {0.00/21 8s (Sb2)} 36. Nb5 {+0.70/18 13s} gxf5 {+0.39/20
11s} 37. gxf5 {+0.75/18 12s} Ne3 {+0.33/19 7s} 38. Nd6+
{+0.78/19 9s} Kg7 {+0.68/20 7s} 39. Kf4 {+0.91/18 9s} Nd1
{+0.73/20 5s} 40. Bd2 {+0.93/18 19s (Sc8)} h5 {+0.80/19 13s
(h6)} 41. h4 {+0.95/18 11s (a4)} Nb2 {+1.26/22 14s} 42. Bc3
{+1.46/19 11s} Nbc4 {+2.10/23 12s (Sbd3+)} 43. Ne8+
{+1.33/20 10s (Sxc4)} Kf7 {0.00/23 12s} 44. Nxf6 {+1.80/21
22s (Sc7)} Nd3+ {+2.60/24 35s} 45. Ke4 {+1.51/20 10s (Kg5)}
Nf2+ {+2.79/21 15s (Sc5+)} 46. Kd5 {+1.90/19 14s} Nxa3
{+2.82/21 10s (Se3+)} 47. Nxh5 {+2.47/17 10s} Nd1 {+3.20/20
11s} 48. Bd2 {+2.47/18 14s (Ld4)} Nc2 {+3.57/19 12s}
49. Ng3 {+2.69/17 11s} Nde3+ {+3.72/21 17s} 50. Ke4
{+2.69/18 12s} Nc4 {+3.71/21 11s (Sg4)} 51. Bg5 {+2.83/17
20s (Lc1)} b5 {+3.38/21 9s} 52. Kd3 {+2.55/17 8s (h5)} Ne1+
{+3.37/20 11s} 53. Kd4 {+3.08/16 8s (Kc3)} Nf3+ {+2.60/20
11s (Sd6)} 54. Kd5 {+2.38/16 15s} Nce5 {+3.02/21 13s}
55. Ke4 {+2.41/19 16s (Se2)} b4 {+3.31/21 11s} 56. Ne2
{+2.34/19 12s} Nxg5+ {+3.16/21 11s (b3)} 57. hxg5 {+2.28/16
1s} Nc4 {+3.07/21 11s} 58. Nd4 {+2.46/18 10s} Nd6+
{+3.41/20 33s (Sa5)} 59. Kf4 {+2.80/19 11s} Kg7 {+3.23/20
10s (Sc4)} 60. Ne6+ {+2.73/18 8s (Sb3)} Kg8 {+3.23/22 10s
(Kh7)} 61. Nc5 {+2.83/20 12s} Kg7 {+3.31/21 9s} 62. Nb3
{+2.89/19 11s} Nf7 {+2.87/23 8s} 63. Nc1 {+2.33/20 11s
(g6)} Nd6 {+2.40/21 10s} 64. Kg4 {+2.22/20 11s} Kf7
{+2.87/22 9s (Sf7)} 65. Nb3 {+2.09/19 9s} Kg7 {+2.87/23
10s} 66. Nd2 {+2.27/20 12s} Kf7 {+2.87/22 9s} 67. Kf4
{+2.61/20 10s} Kg7 {+2.87/22 9s} 68. g6 {+1.54/20 15s} Ne8
{+2.87/23 10s} 69. Kg5 {+1.93/20 11s} Nf6 {+2.87/24 10s}
70. Nb3 {+2.04/20 10s} Ne4+ {+2.87/23 10s} 71. Kf4
{+1.91/19 9s} Nd6 {+3.06/23 10s} 72. Ke5 {+1.80/21 11s}
Nc4+ {+2.44/22 10s} 73. Kd5 {+1.29/20 10s} Ne3+ {+1.57/23
11s} 74. Ke6 {+0.85/21 12s} Ng2 {+1.57/22 9s} 75. f6+
{+1.32/21 10s} Kxg6 {0.00/25 9s} 76. f7 {+1.24/21 12s} Nf4+
{0.00/27 9s} 77. Ke7 {+1.49/20 9s} Kg5 {0.00/27 8s (Kh7)}
78. Nd4 {+0.66/20 16s (Kf8)} Ng6+ {0.00/25 7s} 79. Kd8
{+0.45/20 12s (Ke6)} Kh5 {0.00/24 5s (Kf4)} 80. Ke8
{+0.57/20 14s (Kc8)} Kg4 {0.00/25 17s (Kh6)} 81. Nc6
{+0.07/20 9s} b3 {0.00/24 12s} 82. Ne5+ {+0.51/20 10s} Kg5
{0.00/24 12s} 83. Nxg6 {+0.03/19 9s (Sc4)} b2 {0.00/23 12s}
84. f8=Q {+0.22/22 16s} b1=Q {0.00/34 12s} 85. Qc5+
{0.00/16 10s (Se5)} Kxg6 {0.00/37 14s (Df5)} 86. Ke7
{0.00/16 11s} Qe4+ {0.00/100 0s} 87. Kd6 {0.00/16 10s
(Kd8)} 1/2-1/2
[/pgn]
User avatar
Werner
Posts: 2871
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Werner »

Oops, first Crash with following Windows message:

the program \eoc\pattern.exe cannot be started as it ist not compatible with Windows 64bit. Ask the author, if there is a Version for 64bit Windows too ?

There are 2 different pattern. exe used - and this one is not compatible with Windows 64bit.

So Ed: can you provide a Version which workes on 64bit Windows too like the other pattern.exe ??
...and where comes this ghost pattern.exe ?
Werner
Javier Ros
Posts: 200
Joined: Fri Oct 12, 2012 12:48 pm
Location: Seville (SPAIN)
Full name: Javier Ros

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Javier Ros »

Hello Ed,

I think it's a brilliant and revolutionary idea that will raise the strength of the programs in a significant amount.

Thanks for the contribution.
User avatar
mclane
Posts: 18749
Joined: Thu Mar 09, 2006 6:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub

Re: Rebel TNG (version 1.0) approximately 2950....

Post by mclane »

1. Yesterday was ?
2. See 1.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
User avatar
Werner
Posts: 2871
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Werner »

Werner wrote:Oops, first Crash with following Windows message:

the program \eoc\pattern.exe cannot be started as it ist not compatible with Windows 64bit. Ask the author, if there is a Version for 64bit Windows too ?

There are 2 different pattern. exe used - and this one is not compatible with Windows 64bit.

So Ed: can you provide a Version which workes on 64bit Windows too like the other pattern.exe ??
...and where comes this ghost pattern.exe ?
...as this .exe is gone after klicking on the Windows message, here the whole entry:

c:\Users\xxxx\Arena\Engines\Rebel TNG\Eoc\pattern.exe 337.408 02.04.2014 10:55 -a--
this pattern.exe looks like copied into this folder only short after the opening like here:

[pgn]

[Event "40 Züge in 10 min"]
[Site "ProDeo2, I7"]
[Date "2014.04.02"]
[Round "1.1"]
[White "Rebel TNG"]
[Black "Shredder 12 SE"]
[Result "*"]

1. Nf3 Nf6 2. g3 g6 3. Bg2 Bg7 4. O-O O-O 5. d3 d6 6. e4
Nc6 {0.00/14 7s} 7. Nc3 {+0.01/6 4:50m} e5 {+0.06/15 13s
(Lg4)} *
[/pgn]

Werner
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Rebel »

Werner wrote:I did following experement:
Shredder GUI uses only CPU 3 on my quad and so will do the engines inside this GUI. Then I started a tournment
Rebel TNG - ICE 1.0 x64 1CPU with tbs. The first game ran without Problems. In the taskmanager I saw Rebel.exe and pattern.exe using together the same CPU. Both running with 50% of the CPU when starting - and when pattern.exe is finished Rebel.exe use 100%.
This is indeed the correct behaviour.
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Rebel »

Werner wrote:Oops, first Crash with following Windows message:

the program \eoc\pattern.exe cannot be started as it ist not compatible with Windows 64bit. Ask the author, if there is a Version for 64bit Windows too ?
That's odd, all my stuff is developed (and tested) under W7-64.
There are 2 different pattern. exe used - and this one is not compatible with Windows 64bit.
Nonsense :wink: see above.
So Ed: can you provide a Version which workes on 64bit Windows too like the other pattern.exe ??...and where comes this ghost pattern.exe ?

There is no ghost in your PC, I think it's a crinkle of mine, I will look into it.
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: Rebel TNG (version 1.0) approximately 2950....

Post by Rebel »

Werner wrote:...as this .exe is gone after klicking on the Windows message, here the whole entry:

c:\Users\xxxx\Arena\Engines\Rebel TNG\Eoc\pattern.exe 337.408 02.04.2014 10:55 -a--
this pattern.exe looks like copied into this folder only short after the opening like here:
Nothing is copied :lol:

Hint: do the sizes match?