OliThink 5.4.0 has been published with an big leap in strength for only 3 lines of code

[OliverBr]

On the other hand, when someone does win a game against Magnus Carlsen, people sit up and take notice, even if they know Magnus is a lot better.

It's funny, because 5.6.0 wins a lot of more games against Stockfish 5 than 5.6.1. This verifies my observation that 5.6.0 performs better against other engines, so I recommend 5.6.0 over 5.6.1.

It's like:
5.6.1 > 5.6.0 in h2h games.
5.6.0 > 5.6.1 in games against other engines

[OliverBr]

Version 5.6.5 is finally released. It's a very consistent version with about 150 ELO+ compared to 5.3.3. (now about 2540).
OliThink can beat any opponent, including Stockfish and Leela. (though not very often).

The aim of the project is to increase strength, while not increasing complexity or code size. It's about the same size (1650 lines of code, <100k binary) as Version 5.3.3. Furthermore, there is not even one line of code copied from another engine.

There is now a GIT-project. I committed every version of OliThink 5. More to come (Java, Olithink < 5):

https://github.com/olithink/OliThink/

[Dann Corbit]

That strength rise is SF-NNUE like.
applause

[OliverBr]

That strength rise is SF-NNUE like.
applause

Thank you, but it will get even harder to improve more without expanding the code.
I would like to present a win vs. Stockfish 1. It's a real crazy-tactical game:

[pgn]
[Event "?"]
[Site "?"]
[Date "2020.08.16"]
[Round "670"]
[White "OliThink 5.6.5"]
[Black "Stockfish 11 64"]
[Result "1-0"]
[TimeControl "40/40"]

1. e3 Nf6 2. c3 Nc6 3. b4 e6 4. a4 a6 5. d4 d5 6. Nf3 Bd6 7. Nbd2 e5 8. dxe5
Nxe5 9. Nxe5 Bxe5 10. Bb2 O-O 11. Bd3 Bd6 12. a5 Qe7 13. O-O Rd8 14. Re1 Bxh2+
15. Kxh2 Ng4+ 16. Kg3 h5 17. f4 g5 18. Nf1 Re8 19. Bc2 Qf6 20. Qd4 Re5 21. c4
Bd7 22. cxd5 Rae8 23. Rad1 h4+ 24. Kf3 b6 25. b5 Bc8 26. axb6 cxb6 27. bxa6 Bf5
28. Ba4 Ra8 29. e4 Bd7 30. g3 Bxa4 31. Kxg4 gxf4 32. Qxe5 Qxe5 33. Bxe5 Bxd1+
34. Rxd1 hxg3 35. Kxf4 f6 36. Bxf6 g2 37. Ne3 Kf7 38. e5 b5 39. d6 Ke6 40. Nd5
Rf8 41. a7 g1=Q 42. Nc7+ Kd7 43. Rxg1 Kc6 44. a8=Q+ Rxa8 45. Nxa8 b4 46. Rc1+
Kb5 47. d7 b3 48. d8=Q b2 49. Qd3+ Ka4 50. Qc4+ Ka3 51. Rc3# 1-0
[/pgn]

Stockfish sacrifices the black bishop on move 14:
14...Bxh2+?
[d]r1br2k1/1pp1qppp/p2b1n2/P2p4/1P6/2PBP3/1B1N1PPP/R2QR1K1 b - - 4 14

In move 25, OliThink throws more tactics into the game:
25. b5!
[d]4r1k1/2pb1p2/pp3q2/P2Pr1p1/1P1Q1Pnp/4PK2/1BB3P1/3RRN2 w - - 0 25

[tmokonen]

Am I the only one who has trouble accessing the OliThink website? It appears that my IP has been blocked, since I can connect through a proxy server. I have not been able to connect for a few weeks now.

[NaltaP312]

no problem with vpn for me

[OliverBr]

Am I the only one who has trouble accessing the OliThink website? It appears that my IP has been blocked, since I can connect through a proxy server. I have not been able to connect for a few weeks now.

Fail2ban has banned really a lot of IPs in the last years. I have cleaned it up, can you please check if it works now?

Alternatively you can check OliThink's github site. There are binaries, too. I have just commited 5.6.6:
https://github.com/olithink/OliThink/tree/master/bin

[tmokonen]

Thank you Oliver, I can access your website again.

[OliverBr]

OliThink is approaching ELO 2600. Version 2.6.7 was about 2570 and now 2.6.8 has been released:

Code: Select all

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)     W     D     L  D(%)  CFS(%)
   1 OliThink 5.6.8    :      27     10  1572.5    2920  53.9  1005  1135   780  38.9     100
   2 OliThink 5.6.7    :       0   ----  1347.5    2920  46.1   780  1135  1005  38.9     ---

White advantage = 10.27 +/- 5.22
Draw rate (equal opponents) = 39.09 % +/- 0.89

But there is a catch: This grow is only for Linux and MacOSX due to optimisation, Windows executables have just about 15 ELO gain.

Does anybody know the equivalent for "-mavx2" Option in Visual Studio 2019?

[OliverBr]

Note:
OliThink's purpose is not to become the strongest chess engine. The target is to have a very small and clear code with a minimalistic mobility evaluation.
How strong can such a engine become?