Rhetoric 1.2 Released

[asanjuan]

Hello all.

Rhetoric 1.2 is ready to download on my web page.

www.chessrhetoric.com

I would like to see a lot of downloads. I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tunning) algorithm.

I hope to see comments about his style. I think it plays "human" moves. I love the Rhetorical style.

Cheers.

[Adam Hair]

Thanks, Alberto!

[Graham Banks]

Hello all.

Rhetoric 1.2 is ready to download on my web page.

www.chessrhetoric.com

I would like to see a lot of downloads. I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tunning) algorithm.

I hope to see comments about his style. I think it plays "human" moves. I love the Rhetorical style.

Cheers.

Thanks Alberto.

[carldaman]

Hello all.

Rhetoric 1.2 is ready to download on my web page.

www.chessrhetoric.com

I would like to see a lot of downloads. I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tunning) algorithm.

I hope to see comments about his style. I think it plays "human" moves. I love the Rhetorical style.

Cheers.

Well, thank you, Alberto! I had a feeling you weren't kidding about those 'human' moves, and you are absolutely right! The very first Rhetoric 1.2 test game I ran was a spectacular and crushing win over Ruffian 2.10, whose failure to trade Queens while it had the chance was soundly punished.

Rhetoric's play appears very lively and full of exciting surprises


A very pretty winning position just before the game ended:

[D]6n1/pk3PPp/2p3n1/3p4/8/2QK2P1/6r1/8 b - - 0 43


[pgn]

[Date "2014.02.28"]
[Round "1"]
[White "Rhetoric 1.2_x64"]
[Black "Ruffian_210"]
[Result "1-0"]
[BlackElo "2625"]
[ECO "C18"]
[Opening "French"]
[Time "01:55:22"]
[Variation "Winawer, 6...Qa5"]
[TimeControl "40/600:40/600:40/300"]
[Termination "adjudication"]
[PlyCount "91"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Qa5 7. Bd2 Qa4 8.
Qb1 c4 9. h4 Nc6 10. h5 Bd7 11. Ne2 O-O-O 12. Rh3 {+0.14/17 20} f6
{+0.79/13 13} 13. Rg3 {0.00/17 20} fxe5 {+0.82/13 11} 14. Rxg7 {+0.42/18
20} Nf6 {+0.79/13 21} 15. dxe5 {+0.34/17 20} Nxe5 {+0.59/14 21} 16. h6
{+0.13/16 20} Rdg8 {+0.86/13 16} 17. Nd4 {0.00/15 20} Rxg7 {+0.52/16 72}
18. hxg7 {+0.02/16 20} Rg8 {+0.52/14 19} 19. Qb4 {+0.02/16 20} Ng6
{+0.43/14 19} 20. Qc5+ {+0.33/15 20} Kb8 {+0.84/13 19} 21. Qd6+ {+0.33/15
20} Ka8 {+0.95/14 17} 22. Qc7 {+0.33/16 20} e5 {+0.86/14 19} 23. Rb1
{+1.11/14 20} Rb8 {+0.91/13 23} 24. Bg5 {+1.45/15 20} Ng8 {+0.84/13 17} 25.
Ne2 {+1.45/16 20} Qxc2 {+0.81/13 11} 26. Rd1 {+1.05/16 20} Bc6 {+0.72/13
19} 27. Be3 {+1.43/16 20} Qa4 {+0.80/13 14} 28. g3 {+1.43/16 20} Qxa3
{0.00/15 119} 29. Bg2 {+1.72/15 20} N6e7 {0.00/14 11} 30. Qxe5 {+1.91/16
20} Ba4 {+0.14/12 5} 31. Bc1 {+2.29/15 20} Qb3 {+0.11/13 9} 32. Rd2
{+2.35/14 20} Bc6 {0.00/13 12} 33. Nd4 {+2.10/15 20} Qxc3 {+0.22/14 8} 34.
Nxc6 {+3.14/17 20} Qxc1+ {-0.20/16 13} 35. Rd1 {+3.14/18 20} Qxd1+
{-0.63/16 13} 36. Kxd1 {+3.34/17 20} bxc6 {-0.63/17 13} 37. f4 {+4.38/16
20} Rb1+ {-0.65/15 13} 38. Ke2 {+5.14/16 20} Kb7 {-1.84/14 13} 39. f5
{+5.52/15 20} c3 {-1.85/13 6} 40. Kd3 {+6.17/14 16} Rg1 {-3.02/13 9} 41. f6
{+7.62/14 15} Ng6 {-4.20/14 15} 42. Qxc3 {+8.37/14 15} Rxg2 {-4.44/15 15}
43. f7 {+8.66/15 15} Rxg3+ {-5.55/16 18} 44. Kd2 {+9.27/16 15} Rxc3
{-5.45/15 7} 45. fxg8=Q {+9.23/15 15} Rg3 {-5.78/16 15} 46. Qf7 {+9.33/14
15 User Adjudication} 1-0

[/pgn]

Cheers,
CL

[asanjuan]

You're wellcome!
I always wanted to make something different. I'm glad you like it

[Steve Maughan]

Hi Alberto,

(...). I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tuning) algorithm.(...)

Congrats!

May I ask Which "learning algorithm" you implemented?

I'm a little surprise you attribute the elo increase to (among other things) Singular Extensions. I always thought this added little in terms of raw strength.

Any estimate of the elo gain? I'm interested as I am at the stage of adding these elements to Maverick.

Steve

[Henk]

Hi Alberto,

(...). I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tuning) algorithm.(...)

Congrats!

May I ask Which "learning algorithm" you implemented?

I'm a little surprise you attribute the elo increase to (among other things) Singular Extensions. I always thought this added little in terms of raw strength.

Any estimate of the elo gain? I'm interested as I am at the stage of adding these elements to Maverick.

Steve

If you only check whether the best move from hash is singular, singular extensions will be a special case of a PV extension and LMR never reduces the first move tried. So I don't think that form of singular extensions works well in combination with LMR.

[asanjuan]

Hi Alberto,

(...). I managed to add a lot of elo points with only 3 changes: Singular Extensions, Late Move Pruning, and applying a new learning (or tuning) algorithm.(...)

Congrats!

May I ask Which "learning algorithm" you implemented?

I'm a little surprise you attribute the elo increase to (among other things) Singular Extensions. I always thought this added little in terms of raw strength.

Any estimate of the elo gain? I'm interested as I am at the stage of adding these elements to Maverick.

Steve

SE alone gave me around 20 elo.

LMP was a surprise and gave me 65 elo points!! It was a little bit tricky, because you need the null move threat to avoid tactical mistakes.

The big surprise was the implementation of the tunning based on Peter Osterlünd posts. I implemented the sigmoid function, the average error calculation function based on a big epd file, and the rest was a matter of search. You can use whatever technique to search (gradient, G.A, or a simple loop as Peter pointed). The result is a balanced set of parameters for your evaluation. And it gave Rhetoric another 80 or 90 elo more and a very dynamic and positional style. It discovered stunning values: for example:
A pawn in the opening is worth 75 centipawns, but in the endgame the value rises to 110. The mobility values where doubled and almost doubled again for the endgame except for the queen... That values made rhetoric sacrifice pawns for attack or any other positional criteria.

The total amount added: more than 100 elo points in self play. The test against the pool of sparrings that I have been using until now, show even better results. This may put Rhetoric in a good place in the rating lists. I'm not going to give a number but I have good feelings with this version.

S.E. can be tricky also. I implemented it in a very straightforward way. Quite simple, but working. If you want some code, I can send you an email or a PM with the details. I shared this mail with Brian Richardson (author of Tinker), so is not a problem.


Cheers.

[Steve Maughan]

SE alone gave me around 20 elo.

LMP was a surprise and gave me 65 elo points!! It was a little bit tricky, because you need the null move threat to avoid tactical mistakes.

The big surprise was the implementation of the tunning based on Peter Osterlünd posts. I implemented the sigmoid function, the average error calculation function based on a big epd file, and the rest was a matter of search. You can use whatever technique to search (gradient, G.A, or a simple loop as Peter pointed). The result is a balanced set of parameters for your evaluation. And it gave Rhetoric another 80 or 90 elo more and a very dynamic and positional style. It discovered stunning values: for example:
A pawn in the opening is worth 75 centipawns, but in the endgame the value rises to 110. The mobility values where doubled and almost doubled again for the endgame except for the queen... That values made rhetoric sacrifice pawns for attack or any other positional criteria.

The total amount added: more than 100 elo points in self play. The test against the pool of sparrings that I have been using until now, show even better results. This may put Rhetoric in a good place in the rating lists. I'm not going to give a number but I have good feelings with this version.

S.E. can be tricky also. I implemented it in a very straightforward way. Quite simple, but working. If you want some code, I can send you an email or a PM with the details. I shared this mail with Brian Richardson (author of Tinker), so is not a problem.

Great summary.

I wasn't looking for absolute ELO of new version but the estimated deltas for the elements you added.

Since asking the question I found Peter Osterlünd post. I'll take a look. It looks interesting.

I'm getting bogged down in trying to tune Maverick's evaluation function. The domain space of an evaluation function is *massive*. Auto-tuning must be the way to go.

Thanks,

Steve

[Jamal Bubker]

Thank you Alberto for this new release !