Test no.1 (DiscoCheck w32 comparative)

[Sylwy]

Conditions:

-TC = 4'+2"
-Hash = 64 MB (The Crazy Bishop 0.0.52) and 84 MB (Jazz)
-Books = Perfect_2010.abk
-GUI = Arena 3.0
-CPU = Athlon 64 3800+
-OS = Windows XP SP3

The Crazy Bishop 0.0.52 was linked to Nalimov TBs (+ 16 MB Nalimov cache).



Like marker:



Sorry !

SilvianR

[Michael Sherwin]

Hello CCC folks !



After 200 games/each engine (900 games per total)

1.-Rhetoric 1.4.1...........[86,00%]
2.-The Baron 3.33...........[85,50%]
3.-Hermann 2.8..............[69,00%]
4.-BBChess 1.3b.............[60,50%]
5.-RomiChess P3L............[53,25%]
6.-Bearded 44.5.............[52,50%]
7.-KnightCap 3.7e...........[30,00%]
8.-Neurone XXI..............[ 8,50%]
9.-neuroStock 2.0.1.........[ 4,75%]

The next calculation: after 400 games/each engine (1800 games per total).

Cheers,
SilvianR

RomiChess went from 50% to 53.25% which is not great but considering that the stronger BBChess was added the gain does not seem bad. I know that the first post contained too few games. Looking forward to more results.

[Sylwy]

A nice B class (USCF) chess engine !

SilvianR

[pgn]
[Event "Computer chess game"]
[Site "SILVIAN"]
[Date "2015.03.13"]
[Round "?"]
[White "Rucsandescu Silvian"]
[Black "Byak 8.10.40 JA"]
[Result "1-0"]
[BlackElo "2000"]
[ECO "B13"]
[Opening "Caro-Kann"]
[Time "11:35:57"]
[Variation "Exchange, 4.Nf3 Nf6"]
[WhiteElo "2400"]
[TimeControl "240+2"]
[PlyCount "99"]
[WhiteType "human"]
[BlackType "program"]

1. e4 c6 2. d4 d5 3. exd5 Qxd5 {(Qd8xd5 Nb1-c3 Qd5-a5 Bf1-c4 Qa5-b4
Bc4xf7+) +0.15/6 3} 4. Nf3 Qe4+ {(Qd5-e4+ Bf1-e2 e7-e6 Nb1-c3 Bf8-b4 O-O)
+1.40/6 3} 5. Be2 Bf5 {(Bc8-f5 O-O Bf5-g4 Nb1-c3 Qe4-f5 Bc1-e3) +0.65/6 3}
6. c3 Qg4 {(Qe4-g4 Nf3-g5 Qg4xg2 Be2-f3 Bf5-c2 Bf3xg2) -0.10/6 3} 7. O-O e6
{(e7-e6 Qd1-b3 b7-b6 h2-h3 Qg4-e4) -2.20/6 3} 8. Ne5 Qh4 {(Qg4-h4 Be2-h5
g7-g6 Bh5-f3 Ng8-f6) -2.10/6 3} 9. Nd2 Qd8 {(Qh4-d8 Be2-f3 Bf8-d6 Ne5-c4
Bd6-f4) -2.15/6 3} 10. Ndf3 Be7 {(Bf8-e7 Qd1-b3 Qd8-c8 Bc1-e3 Ng8-f6)
-1.80/6 3} 11. h3 Bf6 {(Be7-f6 Bc1-f4 Nb8-d7 Ne5xd7 Qd8xd7) -1.30/6 3} 12.
b3 Be4 {(Bf5-e4 Bc1-f4 Nb8-d7 Ne5xd7 Qd8xd7) -0.90/6 3} 13. Re1 Bxe5
{(Bf6xe5 Nf3xe5 Be4-d5 Be2-f3 Ng8-f6 Bf3xd5) +0.25/6 3} 14. Nxe5 Nd7
{(Nb8-d7 Ne5xd7 Qd8xd7 Be2-c4 Be4-f5) -1.50/6 4} 15. Nxd7 Qxd7 {(Qd8xd7
Be2-f3 Be4-f5 Bc1-g5 Ra8-b8 Ra1-c1) +0.10/6 4} 16. Bf3 Bg6 {(Be4-g6 Bc1-g5
f7-f6 Bg5-e3 Ra8-d8 Ra1-c1) +0.35/6 4} 17. Ba3 Nf6 {(Ng8-f6 Ba3-c1 h7-h6
Bc1-b2 Ra8-d8 Ra1-c1) +0.45/6 4} 18. Qe2 Rb8 {(Ra8-b8 Qe2-e5 Rb8-d8 Bf3xc6
b7xc6) -1.85/6 4} 19. Rad1 Rd8 {(Rb8-d8 Qe2-e5 h7-h6 Ba3-b2 Nf6-g8) -2.40/6
4} 20. c4 Qc8 {(Qd7-c8 Qe2-e5 h7-h6 Bf3xc6+ b7xc6) -2.30/6 4} 21. Bc5 b6
{(b7-b6 Bc5-a3 Bg6-f5 Ba3-b2 c6-c5 Qe2-f1) +0.40/6 4} 22. Ba3 Nd7 {(Nf6-d7
Qe2-b2 Nd7-b8 Qb2-c1 c6-c5 Ba3-b2) -0.20/6 5} 23. d5 cxd5 {(c6xd5 Ba3-b2
d5-d4 Rd1xd4 e6-e5 Qe2-d1) +0.35/7 5} 24. cxd5 Nc5 {(Nd7-c5 Rd1-c1 Bg6-d3
Qe2-d1 Bd3-e4 Ba3xc5) -0.80/7 5} 25. Bxc5 bxc5 {(b6xc5 d5xe6 Rd8xd1 e6xf7+
Ke8xf7 Re1xd1) -0.95/7 5} 26. dxe6 Rxd1 {(Rd8xd1 e6xf7+ Ke8xf7 Re1xd1
Bg6-f5 Qe2-b2 Bf5-e6) -2.75/7 5} 27. exf7+ Kxf7 {(Ke8xf7 Qe2-e7+ Kf7-g8
Re1xd1 h7-h6 Rd1-d8+ Qc8xd8) -4.00/7 5} 28. Qc4+ Kf8 {(Kf7-f8 Re1xd1 Kf8-e7
Rd1-e1+ Ke7-d8 b3-b4 Bg6-f5) -3.05/7 6} 29. Rxd1 Bf7 {(Bg6-f7 Qc4-h4 Qc8-e6
Rd1-d8+ Bf7-e8 Qh4-f4+) -1.50/6 6} 30. Qf4 Kg8 {(Kf8-g8 Qf4-g5 Qc8-f8
Rd1-d8 Bf7-e8 Bf3-d5+) -1.50/6 6} 31. Bd5 Bxd5 {(Bf7xd5 Rd1xd5 h7-h6 h3-h4
Qc8-c6 Qf4-b8+) -0.70/7 6} 32. Rxd5 a6 {(a7-a6 Qf4-c4 Kg8-f8 Qc4xc5+ Qc8xc5
Rd5xc5) -1.60/7 6} 33. Qd6 Qe8 {(Qc8-e8 Kg1-h2 Qe8-c8 Qd6-d8+ Kg8-f7
Rd5-d7+ Qc8xd7) -3.80/7 7} 34. Qd8 Qf8 {(Qe8-f8 Qd8xf8+ Kg8xf8 Rd5-d8+
Kf8-e7 Rd8xh8 h7-h6) -7.60/7 7} 35. Qxf8+ Kxf8 {(Kg8xf8 Rd5-d8+ Kf8-e7
Rd8xh8 Ke7-f6 a2-a4 h7-h6 f2-f4 Kf6-f5) -8.20/10 8} 36. Rd8+ Ke7 {(Kf8-e7
Rd8xh8 Ke7-f6 Rh8-a8 Kf6-e5 Ra8xa6 Ke5-d4 Ra6-a7) -8.10/9 8} 37. Rxh8 h6
{(h7-h6 Rh8-a8 h6-h5 Ra8xa6 g7-g5 Ra6-c6 Ke7-d7 Rc6xc5 g5-g4 h3-h4)
-8.70/11 8} 38. Ra8 h5 {(h6-h5 Ra8xa6 g7-g5 Ra6-h6 h5-h4 Rh6-g6 Ke7-d7
Rg6xg5 Kd7-c6) -9.10/11 9} 39. Rxa6 Kf7 {(Ke7-f7 Ra6-c6 c5-c4 b3xc4 g7-g5
Kg1-h2 g5-g4 Rc6-c7+ Kf7-e6 h3-h4) -9.50/11 9} 40. Rc6 c4 {(c5-c4 b3xc4
g7-g5 Kg1-h2 Kf7-g7 Kh2-g3 Kg7-h7 Rc6-c7+) -9.85/11 10} 41. Rxc4 g5 {(g7-g5
a2-a4 Kf7-e6 a4-a5 g5-g4 h3xg4 h5xg4 Rc4xg4 Ke6-f6 f2-f4) -10.30/10 11} 42.
a4 Kg7 {(Kf7-g7 a4-a5 Kg7-f6 a5-a6 Kf6-f5 a6-a7 g5-g4 a7-a8Q Kf5-g5
Rc4-c5+) -16.70/11 12 Black resigns}
[/pgn]

[Graham Banks]

A nice B class (USCF) chess engine !

What Elo level is that?

[Michael Sherwin]

A nice B class (USCF) chess engine !

What Elo level is that?

Class D 1200 - 1399
Class C 1400 - 1599
Class B 1600 - 1799
class A 1800 - 1999
expert 2000 - 2199
master 2200 - 2399
senior
master 2400 and above

[Graham Banks]

A nice B class (USCF) chess engine !

What Elo level is that?

Class D 1200 - 1399
Class C 1400 - 1599
Class B 1600 - 1799
class A 1800 - 1999
expert 2000 - 2199
master 2200 - 2399
senior
master 2400 and above

Thanks.

[Sylwy]

Hi !

After 400 games/each engine (1800 games per total):



The situation after 900 games per total (200 games/each engine) was:



Rhetoric 1.4.1 & The Baron 3.33 have very good learning functions:



Take a look how good was the evolution of Rhetoric 1.4.1:



The next publication will be after 2.700 games per total (600/engine) !

SilvianR

[Sylwy]

Hello CCC folks !

After 600 games/engine (2700 games per total) the cross-table is:



After 400 games/engine:



After 200 games/engine:



Like you can see in the below capture, after 2700 games in my tournament, the most effective learning functions seem to be implemented in:

-Bearded 44.5
-RomiChess P3L
-KnightCap 3.7e



The next publication: after 3600 games (800/engine).


Like bonus:

A.- The Institute of Anatomy in Iasi (one of the oldest in Europe):



It is a part of Grigore T. Popa University of Medicine and Pharmacy in IASI - the oldest in Romania !

http://en.wikipedia.org/wiki/Grigore_T. ... d_Pharmacy

B.- Baisa forest in autumn (this forest is not far from Iasi):



Have a nice weekend !

SilvianR

[Michael Sherwin]

The late Marc LaCrosse did an experiment with RomiChess using a huge opening book against about 5 very strong engines. He played a thousand or two thousand games and decided that the learning was not beneficial. I convinced him that Romi's learning would become beneficial shortly if he were to continue. I do not remember how many total games he played. However, RomiChess started to show a learning curve of about +50 elo for every one thousand games played. My own experiment of playing thru a set of 20 positions 10 times against Glaurung 2.x resulted in RomiChess going from 5% to 95% with a couple of setbacks along the way.

Edit: After this experiment is over it would be very interesting to see these engines in a tournament to see how much they gained in elo from this experiment.

[Sylwy]

The late Marc LaCrosse did an experiment with RomiChess using a huge opening book against about 5 very strong engines. He played a thousand or two thousand games and decided that the learning was not beneficial. I convinced him that Romi's learning would become beneficial shortly if he were to continue. I do not remember how many total games he played. However, RomiChess started to show a learning curve of about +50 elo for every one thousand games played. My own experiment of playing thru a set of 20 positions 10 times against Glaurung 2.x resulted in RomiChess going from 5% to 95% with a couple of setbacks along the way.

Edit: After this experiment is over it would be very interesting to see these engines in a tournament to see how much they gained in elo from this experiment.

I'll do it for sure ! I'm extremely interested.

SilvianR