ShashChess

[amchess]

https://github.com/amchess/ShashChess/r ... tag/6.1.3
I completely isolated mcts code.
If there are yet problems on android perhaps it's beyond my current expertise.
As usual, all other versions are welcome.

[amchess]

https://github.com/amchess/ShashChess/r ... /tag/6.2.4

Stockfish patch
Change outposts to single value Feb 8, 2019
Corrected mcts and negative contempt bugs
No skip 25/01/2019 CoffeOne and Corchess (great improvement)

As always, other versions are welcome, in particular, Android and Mac.
Andrea

[Leo]

How do you load shashchess into Arena?

[amchess]

Like any other UCI engine

[amchess]

7.0 released
https://github.com/amchess/ShashChess/releases/tag/7.0

Added match against the latest stockfish:
https://github.com/amchess/ShashChess/wiki/Matches

As usual, the other builds are welcome.
Andrea

[amchess]

I have a proposition.
The goal of ShashChess is to offer a tool for a correspondence player
This is not the goal of Stockfish community.
Now, there are a lot of stockfish derivate.
As far as I know, only my engine has a whole theory behind.
This theory enables a coherent improvement.
For example, on the official stockfish testing framework, there are a lot of yellow patch:
http://tests.stockfishchess.org/tests
Based on my systematic tests, a lot of them can be integrated in the engine with an increasing elo strength in terms of game play and hard positions solving.
Already present results are a lot appreciated:
https://github.com/amchess/ShashChess/w ... -about-him
but the possible improvements are yet a lot!
The problem is I can’t do this alone because I’m a professional programmer and can’t take up a lot of time from my skills improvements also in other important directions.
So, I ask for a help in three directions:
-tests in a systematic fashion like I explained in the wiki:
https://github.com/amchess/ShashChess/wiki/Tests
-discovery of the most promising yellow patches on the framework
-integration of the patches and other eventual ideas.
I think if we want a great tool we have to unify our efforts, if the goal is not the spotlight, but doing something useful: strenght only lies in unity.
Andrea

[peter]

Hi Andrea!
Thanks for the new version!

So, I ask for a help in three directions:
-tests in a systematic fashion like I explained in the wiki:
https://github.com/amchess/ShashChess/wiki/Tests

Downloaded the test positions from there and started with the last one, nr. 326:

[d]r5k1/5p2/4p1n1/6Q1/1p2b3/1P1p3B/2p5/5RK1 b - - 0 1

Why is the given solution 1...Rc8 a single best move?
Black's lost after this one as well as with other moves.

CFish- output after Forward- Backward and 1...Rc8:

2r3k1/5p2/4p1n1/6Q1/1p2b3/1P1p3B/2p5/5RK1 w - - 0 1

Analysis by Cfish 230219 64 POP N:

2.Kh2 Ra8 3.Qe3 Ra2 4.Rc1 Ra5 5.Qxe4 d2 6.Qxc2 dxc1R 7.Qxc1 Rh5 8.Qd2 Rh4 9.Kg3 Re4 10.Bg4 Re5 11.Qxb4 Rg5 12.Kf2 Ne5 13.Be2 Rf5+ 14.Ke1 Nd7 15.Qh4 Rd5 16.b4 f6 17.b5 Kf7 18.Qb4 Nb6 19.Qa5 Nd7 20.Qa7 Kg6 21.b6 Rd6 22.b7 Kf5 23.Kf2 Rd5 24.Qe3
+- (11.09) Depth: 39/67 00:06:09 10851MN, tb=41789278

24 threads of 12x3GHz, 32G hash and 6men Syzygys.
Did the Fish miss something? I can't find a way to hold that.

Alternative moves for Black after 1...Rc8 2.Kh2 with some more Forward- Backward accordingly back to position after
2...Kg7:

2r5/5pk1/4p1n1/6Q1/1p2b3/1P1p3B/2p4K/5R2 w - - 0 1

Analysis by Cfish 230219 64 POP N:

3.Qf6+ Kh6 4.Qd4 c1Q 5.Rxc1 Rc2+ 6.Rxc2 dxc2 7.Qe3+ Kg7 8.Bg2 Bf5 9.Kg3 Kf6 10.Qh6 Ke7 11.Qd2 e5 12.Qxb4+ Kd7 13.Qb5+ Ke7 14.Qc5+ Kf6 15.Qd6+ Kg7 16.Qd2 Nf4 17.Bf1 Be4 18.Bc4 Nh5+ 19.Kf2 Nf6 20.Ke2 Bg6 21.Bd3 e4 22.Bc4 Bh5+ 23.Ke1 Nh7 24.Qxc2 Ng5 25.Qc3+ Kg6 26.Qb4 Bf3 27.Kd2 f6 28.Be2 Bh1 29.Bf1 Nf3+ 30.Ke3 Ne5 31.Be2 Kf7 32.Qd6 Kg6 33.Qe7 Bg2 34.Qf8
+- (11.25) Depth: 36/67 00:00:50 1480MN, tb=6080054

2...Rf8:

5rk1/5p2/4p1n1/6Q1/1p2b3/1P1p3B/2p4K/5R2 w - - 0 1

Analysis by Cfish 230219 64 POP N:

3.Rg1 Bd5 4.Qe3 Bxb3 5.Qxd3 Rc8 6.Rc1 Rc3 7.Qd6 Rc4 8.Qg3 Ba2 9.Qd3 Rc3 10.Qd8+ Kh7 11.Qa5 Bd5 12.Qxb4 Rc4 13.Qb2 Be4 14.Qf6 Rc7 15.Qd4 Bf3 16.Qd8 Rc4 17.Qd3 Bd5 18.Rxc2 Rh4 19.Rf2 f5 20.Qc3 f4 21.Qf6 f3 22.Kg3 Rf4 23.Qg5 Ra4 24.Bg4 Ra1 25.Rh2+ Kg7 26.Rc2 e5
+- (11.41) Depth: 39/75 00:06:44 11298MN, tb=72429593

Tried to let ShashChess play out with Black after 1...Rc8 against CFish as well, there isn't any way to build up a fortress and prevent the 1-0, is there?

[amchess]

Hi Peter.

I don't know.
It's a collection of positions from the net, for example, here.
So, I didn't check them one by one.
Perhaps, there is an error.
I have to control.

Best regards,
Andrea

[Shahrad]

[Event "Setting ICC Analysis to 8"]
[Site "?"]
[Date "2019.03.04"]
[Round "?"]
[White "Position 326 Sugar 1.6.1 BMI2"]
[Black "Unchecking Null Move" "Unchecking Dynamic Strategy"]
[Result "*"]
[Annotator "Shahrad"]
[SetUp "1"]
[FEN "r5k1/5p2/4p1n1/6Q1/1p2b3/1P1p3B/2p5/5RK1 b - - 0 1"]
[PlyCount "82"]
[SourceVersionDate "2019.03.04"]

{SugaR 1.6.1 64 BMI2:} 1... Rc8 2. Kh2 Kg7 3. Qf6+ Kh6 4. Qd4 f5 5. Rg1 Rc7 6.
Qf6 Rg7 7. Qxe6 d2 8. Bxf5 Bxf5 9. Qe3+ Kh7 10. Qxd2 Ne5 11. Rxg7+ Kxg7 12.
Qg5+ Ng6 13. Kg3 Be6 14. Kf2 Bxb3 15. Qc5 Kf6 16. Qc6+ Kg7 17. Ke3 Ne5 18. Qg2+
Kf6 19. Kd2 Nc4+ 20. Kc1 Ne5 21. Qe4 Ke6 22. Kd2 Bd5 23. Qxc2 {[%eval 98,27]} {
SugaR 1.6.1 64 BMI2:} Kd6 24. Qc8 Nc6 25. Qf8+ Kc7 26. Qc5 Be6 27. Kc1 Bb3 28.
Kb1 Ba4 29. Qg1 Kd6 30. Kb2 Kd5 31. Qa1 Bb5 32. Qh1+ Kc5 33. Qh5+ Kc4 34. Qf7+
Kc5 35. Qf8+ Kd4 36. Qf4+ Kc5 37. Qg5+ Kc4 38. Qc1+ Kd5 39. Kb3 Be2 40. Qg5+
Kd6 41. Ka4 Bc4 42. Qh5 {[%eval 168,27]} *

It finds this solution after 19 half moves. I am not an engine pro but I assume that by hard positions you usually cannot find the solution using Null Move Heuristic.
Black needs two minor pieces plus one pawn vs Queen.

[Shahrad]

Apparently Sugar's solution doesn't work too. I have only a 4 year old i3 and 16gb ram. As far as I checked its analysis, there are several improvements for White in the first 12 half moves.