TalkChess.com

Computer Chess Club
https://talkchess.com/

Bright 0.2c freeware blitz test

https://talkchess.com/viewtopic.php?t=18573

Page 1 of 1

Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 3:22 am
by Mike S.
5m+2s, Athlon@1500 (Win98)
Arena 1.1, Xmas2640.abk
128 MB hash each
ponder off, 32 MB tbs. cache
3/4 + R-R-5-piece Nalimovs

(Bright with 3/4-piece bitbases)

Code: Select all

Bright-0.2c    2690 - Alaric 707         2750   2.0 - 4.0    +1/-3/=2    33.33%
Bright-0.2c    2690 - Colossus 2007d     2680   4.5 - 1.5    +3/-0/=3    75.00%
Bright-0.2c    2690 - Delfi 5.2          2670   4.5 - 1.5    +4/-1/=1    75.00%
Bright-0.2c    2690 - Glaurung 2.0.1     2770   2.0 - 4.0    +0/-2/=4    33.33%
Bright-0.2c    2690 - Naum 2.0           2760   4.0 - 2.0    +3/-1/=2    66.67%
Bright-0.2c    2690 - Pro Deo 1.5        2690   3.0 - 3.0    +2/-2/=2    50.00%
Bright-0.2c    2690 - Ruffian 1.0.5      2660   2.0 - 4.0    +2/-4/=0    33.33%
Bright-0.2c    2690 - Spike 1.2 Turin    2810   2.5 - 3.5    +1/-2/=3    41.67%
Bright-0.2c    2690 - Toga II 1.3.4      2850   0.0 - 6.0    +0/-6/=0    0.00%
Bright-0.2c    2690 - Yace 0.99.87       2570   3.0 - 3.0    +3/-3/=0    50.00%
.
Bright-0.2c    total score 27.5/60 (45.83%); oppo.average ~2720*
*) The ratings of 2690 above means the performance in this test only, relative to the opponent's ratings which are averages from CCRL and CEGT (blitz).

A nice win (see 22.Rxf6!):

[Event "5m+2s Athlon@1500 MHz"]
[Site "Schrotty"]
[Date "2007.12.26"]
[Round "3"]
[White "Bright-0.2c"]
[Black "Delfi 5.2"]
[Result "1-0"]
[ECO "E15"]
[WhiteElo "2690"]
[BlackElo "2670"]
[PlyCount "50"]

1. d4 e6 2. c4 Nf6 3. Nf3 b6 4. g3 Ba6 5. Qa4 Bb7 6. Bg2 c5 7. dxc5 Bxc5 8. O-O
O-O 9. Nc3 Be7 10. Rd1 Bc6 11. Qc2 Na6 12. a3 Rc8 13. e4 Qc7 14. Nd4 Ba8 15.
Bf4 Qc5 16. Rac1 e5 17. b4 Qd6 18. Nf5 Qe6 19. Bg5 Bd8 20. Bxf6 Bxf6 21. Rd6
Qe8 22. Rxf6 gxf6 23. Qd2 Qe6 24. Qh6 Qxf5 25. exf5 Bxg2 1-0

Problems after a (too short) book variation: No 7.axb4, but exchange of the fianchetto bishop on c6 in the next move:

[Event "5m+2s Athlon@1500 MHz"]
[Site "Schrotty"]
[Date "2007.12.26"]
[Round "1"]
[White "Bright-0.2c"]
[Black "Yace 0.99.87"]
[Result "0-1"]
[ECO "A37/27"]
[WhiteElo "2690"]
[BlackElo "2570"]
[PlyCount "77"]

1. c4 c5 2. Nc3 g6 3. g3 Bg7 4. Bg2 Nc6 5. a3 e6 6. b4 Nxb4 7. Rb1 Nc6 8. Bxc6
dxc6 9. Ne4 b6 10. Nf3 f5 11. Neg5 h6 12. Nh3 g5 13. Nhg1 g4 14. Nh4 Ne7 15.
Qc2 Rf8 16. h3 Rf7 17. Rb3 Qd6 18. a4 Bb7 19. Ng2 O-O-O 20. a5 b5 21. hxg4 fxg4
22. Ne3 Nf5 23. Nxf5 exf5 24. Bb2 Bxb2 25. Qxb2 bxc4 26. Rc3 f4 27. Rh4 fxg3
28. Rxg3 Ba6 29. Qc1 Qf6 30. Rhxg4 Qxf2+ 31. Kd1 Bb5 32. Kc2 Rb7 33. Kb2 c3+
34. Kxc3 Qf6+ 35. Kc2 c4 36. Kb1 Ba4+ 37. Ka2 Bb3+ 38. Ka3 Rd5 39. Rg8+ 0-1

Typical Bright depths in these games were 10...12 plies, sometimes bigger.

Games with engine infos:

http://static.twoday.net/computerschach ... artien.zip (47 KB)

Re: Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 10:40 am
by GS
Mike S. wrote:5m+2s, Athlon@1500 (Win98)
Arena 1.1, Xmas2640.abk
128 MB hash each
ponder off, 32 MB tbs. cache
3/4 + R-R-5-piece Nalimovs

(Bright with 3/4-piece bitbases)

Code: Select all

Bright-0.2c    total score 27.5/60 (45.83%); oppo.average ~2720*
...

Problems after a (too short) book variation: No 7.axb4, but exchange of the fianchetto bishop on c6 in the next move:

[Event "5m+2s Athlon@1500 MHz"]
[Site "Schrotty"]
[Date "2007.12.26"]
[Round "1"]
[White "Bright-0.2c"]
[Black "Yace 0.99.87"]
[Result "0-1"]
[ECO "A37/27"]
[WhiteElo "2690"]
[BlackElo "2570"]
[PlyCount "77"]

1. c4 c5 2. Nc3 g6 3. g3 Bg7 4. Bg2 Nc6 5. a3 e6 6. b4? 0-1
...
[D] r1bqk1nr/pp1p1pbp/2n1p1p1/2p5/1PP5/P1N3P1/3PPPBP/R1BQK1NR b KQkq b3 0 6

I wouldn't say the problem here was the too short book line, but that
6.b4(from book) is simply bad for engine games and will always lead
to a score >75% for Black! between programs in a similar range
of ratings. No program would ever play 6.b4 if it has to think about
the position. Humans may handle that line better, but I doubt it.

Guenther

Re: Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 12:06 pm
by Mike S.
GS wrote: I wouldn't say the problem here was the too short book line, but that
6.b4(from book) is simply bad for engine games and will always lead
to a score >75% for Black! between programs in a similar range
of ratings.
Yes, meanwhile I saw that this line is more complicated than I thought at first (some continuations are adventurous, compensation versus material) and most certainly not suitable for computer chess. In the Fritz 7 book, 6.b4 is set to "?" and "no tournament move".

I looked it up in Shredder's online opening databases, and - among human players - White's score after 6.b4 Nxb4 is 50%, but the Elo average is 2523 only. All replied 7.axb4. - After trying engines with this, it seems to me that after 6.b4 Nxb4 7.axb4 cxb4 8.d4 bxc3, 9.Qd3 is considerable... but I am not sure if it's good or if it changes much.

Re: Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 2:08 pm
by Dr.Wael Deeb
GS wrote:
Mike S. wrote:5m+2s, Athlon@1500 (Win98)
Arena 1.1, Xmas2640.abk
128 MB hash each
ponder off, 32 MB tbs. cache
3/4 + R-R-5-piece Nalimovs

(Bright with 3/4-piece bitbases)

Code: Select all

Bright-0.2c    total score 27.5/60 (45.83%); oppo.average ~2720*
...

Problems after a (too short) book variation: No 7.axb4, but exchange of the fianchetto bishop on c6 in the next move:

[Event "5m+2s Athlon@1500 MHz"]
[Site "Schrotty"]
[Date "2007.12.26"]
[Round "1"]
[White "Bright-0.2c"]
[Black "Yace 0.99.87"]
[Result "0-1"]
[ECO "A37/27"]
[WhiteElo "2690"]
[BlackElo "2570"]
[PlyCount "77"]

1. c4 c5 2. Nc3 g6 3. g3 Bg7 4. Bg2 Nc6 5. a3 e6 6. b4? 0-1
...
[D] r1bqk1nr/pp1p1pbp/2n1p1p1/2p5/1PP5/P1N3P1/3PPPBP/R1BQK1NR b KQkq b3 0 6

I wouldn't say the problem here was the too short book line, but that
6.b4(from book) is simply bad for engine games and will always lead
to a score >75% for Black! between programs in a similar range
of ratings. No program would ever play 6.b4 if it has to think about
the position. Humans may handle that line better, but I doubt it.

Guenther
I agree all the way long with you Guenther,even the humans will slip into a bad position no matter how many Elo they have....

Re: Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 3:08 pm
by swami
Mike S. wrote: Yes, meanwhile I saw that this line is more complicated than I thought at first (some continuations are adventurous, compensation versus material) and most certainly not suitable for computer chess.
I think human players have had pretty good stats playing Benko gambit which follows the similar route as this.

it is essential that one need to play the same line a lot and study the games in depth in order to grasp the strength and value of that line.

In computer chess, it may not help as the material values are more important when an engine at 3000 level plays equally stronger engine...

Re: Bright 0.2c freeware blitz test

Posted: Thu Dec 27, 2007 4:39 pm
by GS
swami wrote:
Mike S. wrote: Yes, meanwhile I saw that this line is more complicated than I thought at first (some continuations are adventurous, compensation versus material) and most certainly not suitable for computer chess.
I think human players have had pretty good stats playing Benko gambit which follows the similar route as this.

it is essential that one need to play the same line a lot and study the games in depth in order to grasp the strength and value of that line.

In computer chess, it may not help as the material values are more important when an engine at 3000 level plays equally stronger engine...
Well, it is ... not the Benko/Wolga gambit we are talking about here.
I disagree about similar plans for both, this one here has not much
except b4/b5 in common with it.

Guenther

Re: Bright 0.2c freeware blitz test

Posted: Fri Dec 28, 2007 7:34 am
by Dann Corbit
Dr.Wael Deeb wrote:
GS wrote:
Mike S. wrote:5m+2s, Athlon@1500 (Win98)
Arena 1.1, Xmas2640.abk
128 MB hash each
ponder off, 32 MB tbs. cache
3/4 + R-R-5-piece Nalimovs

(Bright with 3/4-piece bitbases)

Code: Select all

Bright-0.2c    total score 27.5/60 (45.83%); oppo.average ~2720*
...

Problems after a (too short) book variation: No 7.axb4, but exchange of the fianchetto bishop on c6 in the next move:

[Event "5m+2s Athlon@1500 MHz"]
[Site "Schrotty"]
[Date "2007.12.26"]
[Round "1"]
[White "Bright-0.2c"]
[Black "Yace 0.99.87"]
[Result "0-1"]
[ECO "A37/27"]
[WhiteElo "2690"]
[BlackElo "2570"]
[PlyCount "77"]

1. c4 c5 2. Nc3 g6 3. g3 Bg7 4. Bg2 Nc6 5. a3 e6 6. b4? 0-1
...
[D] r1bqk1nr/pp1p1pbp/2n1p1p1/2p5/1PP5/P1N3P1/3PPPBP/R1BQK1NR b KQkq b3 0 6

I wouldn't say the problem here was the too short book line, but that
6.b4(from book) is simply bad for engine games and will always lead
to a score >75% for Black! between programs in a similar range
of ratings. No program would ever play 6.b4 if it has to think about
the position. Humans may handle that line better, but I doubt it.

Guenther
I agree all the way long with you Guenther,even the humans will slip into a bad position no matter how many Elo they have....
The Rybka 2.3.2a analysis of this position shows Rybka thinks it is about even. The choice of Nxb4 was a bit counter-intuitive for me.

Code: Select all

Analysis from c:\epd\tragic.epd   
12/27/2007 10:20:38 PM Level: 385 Seconds
Analyzing engine: Rybkav2.3.2a.w32

1) b3 0 6               
    Avoid move: 
    Best move (Rybkav2.3.2a.w32): Nc6xb4
    Not found in: 06:25
      5	00:00	       4.364	31.469	+0.12	c5xb4 a3xb4 Ng8e7
      6	00:00	      10.405	39.905	+0.09	c5xb4 a3xb4 Nc6xb4 Bc1a3 Qd8b6
      6	00:00	      14.021	43.639	+0.14	Ng8e7 Ng1f3 c5xb4
      7	00:01	      22.575	35.185	+0.09	Ng8e7 b4xc5 Qd8a5 Ng1f3
      7	00:01	      28.458	39.593	+0.24	Nc6xb4 Ra1b1 Nb4c6 Ng1h3
      8	00:01	      38.783	43.785	+0.27	Nc6xb4 Ra1b1 Nb4c6 Ng1f3 Ng8f6
      9	00:01	      72.031	49.636	+0.11	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Qd8f6 Bc1d2 Ng8e7
     10	00:02	     117.517	52.366	+0.09	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Qd8f6 Bc1d2 Ng8e7 Ng1f3
     11	00:04	     190.034	53.211	+0.14	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Ng8f6 Ng1f3 00 00 Nf6e4
     12	00:08	     443.909	59.482	+0.21	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Ng8e7 Ng1f3 d7d5 Nf3d4 00 00
     13	00:14	     800.074	59.174	+0.16	Nc6xb4 a3xb4 c5xb4 Bc1b2 b4xc3 Bb2xc3 Bg7xc3 d2xc3 Ng8f6 Ng1f3 00 00 Qd8b6 Qd1d3
     14	00:22	   1.272.002	58.868	+0.15	Nc6xb4 a3xb4 c5xb4 Bc1b2 b4xc3 Bb2xc3 Bg7xc3 d2xc3 Ng8f6 Ng1f3 00 00 Qd8b6 Qd1d3
     15	00:38	   2.231.372	59.832	+0.16	Nc6xb4 a3xb4 c5xb4 Bc1b2 b4xc3 Bb2xc3 Bg7xc3 d2xc3 Ng8e7 Ng1f3 00 00 Qd8c7 Qd1d3
     16	01:26	   5.169.563	61.642	+0.20	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Ng8e7 Ng1f3 d7d5 Nf3d4 00 c4xd5 Ne7xd5 Bc1b2
     17	02:33	   9.204.400	61.489	+0.15	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Ng8e7 Ng1f3 d7d5 00 Qd8c7 c4xd5 Ne7xd5 Bc1b2
     18	04:57	  17.878.319	61.559	+0.19	Nc6xb4 a3xb4 c5xb4 Qd1b3 b4xc3 d2xc3 Ng8e7 Ng1f3 d7d5 00 Qd8c7 c4xd5 Ne7xd5 Bc1b2
   12/27/2007 10:27:06 PM, Time for this analysis: 00:06:25, Rated time: 06:25

0 of 1 matching moves
12/27/2007 10:27:07 PM, Total time: 12:06:28 AM
Rated time: 06:25 = 385 Seconds
SCID openings report:

Code: Select all

4.2 Moves from the report position

    Move   ECO       Frequency    Score  AvElo Perf AvYear %Draws
 1: Nxb4              84: 48.5%   45.8%              1998   27%
 2: d6                39: 22.5%   57.6%              1999   28%
 3: cxb4              36: 20.8%   77.7%              1996   22%
 4: Nge7              13:  7.5%   73.0%              1993   31%
 5: [end]              1:  0.5%   50.0%              1999  100%
_______________________________________________________________
TOTAL:               173:100.0%   57.2%              1997   27%
All times are UTC+02:00
Page 1 of 1

Powered by phpBB® Forum Software © phpBB Limited