Can a human still defeat Leela ?

[JohnS]

Nice wins Jhor!

The best I can do at the moment is beat it on easy (net 137). Leela didn't do anything about my connected passed pawns.

[pgn]

1. e4 e5
2. Nf3 Nc6
3. Bc4 Nf6
4. d3 d5
5. exd5 Nxd5
6. O-O Be7
7. Re1 O-O
8. Nxe5 Nxe5
9. Rxe5 Nb6
10. Bb3 Bf6
11. Re1 a5
12. c3 a4
13. Bc2 Bf5
14. d4 Bxc2
15. Qxc2 Re8
16. Be3 Nd5
17. Nd2 Nxe3
18. Rxe3 Rxe3
19. fxe3 Bg5
20. Re1 Qd7
21. e4 c5
22. d5 Bxd2
23. Qxd2 Re8
24. c4 b5
25. b3 axb3
26. axb3 bxc4
27. bxc4 Qd6
28. h3 Qe5
29. Qe3 f5
30. Qxc5 fxe4
31. Qe3 h6
32. Rd1 Rc8
33. c5 Rd8
34. d6 Rc8
35. d7 Rd8
36. c6 Qc7
37. Rc1 Kh8
38. Qxe4 Qb6+
39. Kh1 Qc7
40. Qe8+ Kh7
41. Qxd8 Qxd8
42. c7 Qxd7
43. c8=Q Qxc8
44. Rxc8 Kg6
45. Rc5 Kf6
46. Kg1 Ke7
47. Kf2 Kd6
48. Ra5 Ke7
49. Ke3 Kf6
50. Kf4 g5+
51. Kg4 Kg6
52. Ra6+ Kg7
53. Kh5 Kf7
54. Kxh6 Ke7
55. Kxg5 Kd7
56. h4 Kc8
57. Ra7 Kb8
58. Rf7 Kc8
59. h5 Kb8
60. h6 Kc8
61. h7 Kb8
62. h8=Q#
[/pgn]

[fern]

Wait, wait...I have got draws playing against HARD Leela.....

Mmmm

Maybe I am better than I think regards
Fern

[Father]

Thanks Leo,

[Jhoravi]

Nice wins Jhor!

The best I can do at the moment is beat it on easy (net 137). Leela didn't do anything about my connected passed pawns.

I beat Hard Mode again at ID 252. The downgrade to 400 nodes hurt its tactics too much.
[pgn]
1. c4 e5
2. Nc3 Nf6
3. g3 Bb4
4. Bg2 O-O
5. e4 Bxc3
6. bxc3 d6
7. Ne2 Nbd7
8. h3 b6
9. O-O Re8
10. d3 h6
11. f4 Bb7
12. g4 Nh7
13. Ng3 exf4
14. Bxf4 Nhf8
15. Qd2 Ng6
16. Bxh6 gxh6
17. Qxh6 Re5
18. Rxf7 Kxf7
19. Qh7+ Ke6
20. d4 Qg5
21. d5+ Kf6
22. Nh5+ Qxh5
23. gxh5 Ndf8
24. Qxc7 Re7
25. Qxd6+ Kg7
26. Rf1 Rf7
27. Rxf7+ Kxf7
28. hxg6+ Nxg6
29. Qd7+ Kf6
30. Qxb7 Rd8
31. Qc6+ Kg5
32. Qe6 Rf8
33. e5 Nf4
34. Qg4+ Kh6
35. e6 Ng6
36. e7 Nxe7
37. Qh4+ Kg6
38. Qxe7 Rf7
39. Be4+ Kg7
40. Qe5+ Kf8
41. d6 Rg7+
42. Kf2 Rf7+
43. Ke3 Rg7
44. Qf6+ Kg8
45. Bd5+ Kh7
46. d7 Rxd7
47. Qf5+ Kh8
48. Qxd7 a6
49. Be4 Kg8
50. Bg6 b5
51. Qf7+ Kh8
52. Qh7#
[/pgn]

[rainhaus]

All in the title. Is it already too late to beat Leela as a human ? Seems the leela bot on lichess loose only against others bot...

About home tournaments, zero chess and predictions

Autodidactic or self-referential machine chess is still so new and unexplored that medicine men and fortune tellers are probably the best predictors at the moment. Significant longitudinal comparisons over a longer period of time are still completely missing. I don't have the slightest idea how strong the autodidacts really are and can still become. As usual, the numerous match results from private hobby test laboratories can hardly be brought down to a common denominator. The time control, the settings and the hardware are too different. As long as no valid learning curve up to the so-called learning plateaus is visible for the Zeros, an empirically founded prognosis from when homo sapiens leaves the arena is not possible anyway.
Sure, a forum is not a scientific institution. I wouldn't like to miss my exciting home tournaments despite the lack of standards and sufficient quality control. I only bring it up because I have the impression with some questions and presentations that something like a critical background consciousness is not recognizable.

Another significant question is whether the human/LCZero tournaments and games take place in public or are in-house. With the computer alone, it is not hard to get on the throne

Even the quick-witted experts vary between heaven and hell, euphoria and scepticism. Opinions range from extraterrestrial (Magnus' Carlsen's second Peter Heine-Nielsen) and world-changing (chess.com), to human-like B-style (G.Kasparov), constricting like Nine Men's Morris (Th. Czub). At the other end of the scale is the criticism that the cult duel Stockfish/AlphaChessZero is completely overrated because of unfair competition conditions (T. Romstad, H. Nakamura, L.Kaufmann)

Anyhow, looks almost as if a new chess calculation system without classical search algorithms, proven heuristics and traditional evaluation parameters, but with the help of neural networking and Monte Carlo simulation, gave the outprogrammed computer chess a powerful boost. Perhaps synthesis and alpha-beta-NN hybrids will be the result!

As a computer chess veteran I look forward to the further development of the ZERO projects with relaxed eagerness until the chaff separates from the wheat and the contours and limits of the self-feeding chess entity become visible. When Leela has climbed the first TCEC-3000 Elo peak, I also load the Bot into my Ryzen, as GPU version, of course.

On the subject of expert opinions, an interesting forecast from 1991.
The question was: When does a computer defeat the world champion? The list of experts is impressive and the variance of estimates is also....:

1992 Monroe Newborn
1993 John McCarthy
1994 Hans Berliner, Marty Hirsch, Feng-hsiung Hsu
1995 Murray Campbell, Dave Kittinger, David Slate, Mike Valvo
1999 Claude Shannon, Frederic Friedel
2000 Robert Hyatt, Jaap van den Herik
2001 John Nunn.
2001 Julio chaplain
2005 Richard Lang
2008 Harry Nelson
2010 Gary Kasparov ("or maybe never")
2012 Dieter Steinwender
2014 David Levy
2018 Ken Thompson
2020 Tony Marsland
2030 Frans Morsch
2040 Jonathan Schaeffer
from: Chess on the PC, Steinwender/Friedel 1995

Now you just need to know when this event happened
rn

[Jhoravi]

After a long time I tried Hard Mode again and I still beat it. This time against id330. I'm playing white.

https://lichess.org/MkkIaRWp#0