Testing the new Lc0

at least in the first games the search depths were too low for this time control, now it seems to work again.

[pgn][Event "Lc0-v28-rubi"]
[Site "EDDY-PC"]
[Date "2021.08.31"]
[Round "116"]
[White "RubiChess_2.2"]
[Black "Lc0-v28"]
[Result "0-1"]
[BlackElo "2200"]
[ECO "D10"]
[Opening "Slav"]
[Time "08"]
[Variation "Winawer Countergambit Accepted"]
[WhiteElo "2200"]
[TimeControl "40/120:40/120:40/120"]
[Termination "adjudication"]
[PlyCount "99"]
[WhiteType "program"]
[BlackType "program"]

1. e4 e5 2. f4 exf4 3. Nc3 Qh4+ 4. Ke2 d5 5. Nxd5 Bg4+ 6. Nf3 Bd6 7. d4
Bxf3+ {+0.03/10 6} 8. gxf3 {+0.38/24 5} c6 {+0.07/11 1} 9. Nc3 {+0.10/24 4}
Ne7 {+0.09/11 5} 10. Qe1 {-0.12/23 5} Qh5 {+0.10/12 1} 11. Kd1 {-0.16/24 2}
Nd7 {+0.37/8 3} 12. Qf2 {-0.09/22 3} O-O {+0.37/10 4} 13. Ne2 {-0.32/23 8}
Ng6 {+0.34/11 3} 14. c3 {-0.20/24 2} Rfd8 {+0.31/12 10} 15. Bd2 {-0.40/25
5} c5 {+0.31/12 1} 16. Rc1 {-0.31/25 4} b5 {+0.33/1 0} 17. Kc2 {-0.27/25 6}
Nb6 {+0.38/10 2} 18. Kb1 {-0.43/23 4} Nc4 {+0.37/11 1} 19. Be1 {-0.50/23 3}
Be7 {+0.62/9 3} 20. h4 {-0.71/23 2} a5 {+0.62/7 1} 21. Rh3 {-0.83/23 7}
Rac8 {+0.81/7 5} 22. Rh2 {-0.67/24 5} Rb8 {+0.84/7 2} 23. Rc2 {-0.74/23 4}
Bd6 {+1.33/8 3} 24. Bh3 {-0.94/22 5} Ne3 {+1.62/9 2} 25. Bg4 {-1.58/21 2}
Qh6 {+1.46/11 1} 26. Rc1 {-1.44/20 2} Ne5 {+2.99/10 2} 27. dxe5 {-2.73/20
3} Bxe5 {+3.17/8 1} 28. Rc2 {-3.12/23 3} Nxc2 {+2.77/10 6} 29. Kxc2
{-5.07/23 3} Qc6 {+3.02/10 1} 30. Kb1 {-4.90/24 1} h5 {+3.75/10 5} 31. Bxh5
{-4.60/21 3} b4 {+3.59/14 2} 32. cxb4 {-4.91/21 2} cxb4 {+3.56/15 3} 33.
Nc1 {-5.60/26 3} Rbc8 {+5.63/13 3} 34. Qc2 {-3.88/25 2} Qh6 {+7.74/13 2}
35. Bxf7+ {-4.28/24 1} Kxf7 {+7.38/12 1} 36. Qa4 {-4.42/25 1} Qb6 {+7.80/8
10} 37. Re2 {-4.10/24 1} Kg8 {+17.51/6 1} 38. Bf2 {-5.09/24 1} Qa6
{+10.25/8 11} 39. Rc2 {-4.74/23 0} Kh7 {+11.16/9 2} 40. Bc5 {-5.15/19 0}
Rd1 {+11.85/7 0} 41. Qb3 {-6.35/26 4} Rc7 {+8.50/9 16} 42. Qa4 {-6.38/27 4}
Rcd7 {+6.84/8 10} 43. Qb3 {-6.87/26 4} a4 {+16.00/9 9} 44. Qc4 {-7.44/29 4}
Qxc4 {+29.78/10 3} 45. Rxc4 {-7.76/27 2} b3 {+39.65/9 3} 46. axb3 {-8.03/28
4} axb3 {+34.49/8 6} 47. Rc3 {-8.80/28 4} Bxc3 {+57.30/8 7} 48. bxc3
{-9.15/27 2} R7d2 {+53.52/7 5} 49. Ba3 {-11.75/28 5} Rf1 {+64.06/6 7} 50.
h5 {-10.41/26 4  Arena Adjudication} 0-1[/pgn]

Strangling the opponent in Nimzovitch style.

Note 26...Ne5 over and out.

While Lc0 - Rubichess is still in progress curiosity got the upperhand and I made a temporarily rating list.

Code:

   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)
   1 Stockfish 14           :  3685.8   13.3  3248.5    4100    79      98  2487  1523    90    37
   2 Lc0 v28                :  3587.6   12.3   538.0     921    58      52   303   470   148    51
   3 Komodo-Dragon 2        :  3587.2   14.9  3252.0    4500    72      69  2298  1908   294    42
   4 Ethereal 13.25-NNUE    :  3463.4    7.1  1865.5    3400    55      99  1055  1621   724    48
   5 SlowChess 2.7          :  3433.8   11.3   995.5    2200    45      98   411  1169   620    53
   6 RubiChess 2.2          :  3407.1    8.6  1588.5    3321    48      88   795  1587   939    48

So far Lc0 surpassed the mighty Dragon with only 0.4 elo

In my tournament lc0 0.28.0 has so far played 213 games.
So far this is rank 20 from 38.
It got 184 from 213.

Leader in the tournament is lc0 0.27.0 with 311/361
Stockfish 14 (8 cores) 304/362
Ethereal 13 (8 cores) 278/362
Dragon (8 cores) 267/361
Wasp4.50 ( 221/361
Deephiarcs14( 183,5/362
Prodeo3.1 119/361
Prodeo YAT 106,5/361
Colossus2008b 84/362


Lc0 versions on a nvidia2070 using j92-100 net.
All normal engines using 8 cores with permanent brain on a ryzen9/3950x.

I don't see J92-100 in the best nets list.

https://lczero.org/play/networks/bestnets/

J92-100 is a very old net, trained bei Jhorthos https://github.com/jhorthos/lczero-training/wiki/Leela-Training

Code:

Name                Gms  :  Win  : Draw  : Lose  :  Pts  
Lc0-v28             200  :  93+  :  94=  :  13-  : 140.0
RubiChess_2.2       200  :  13+  :  94=  :  93-  :  60.0

Also 70%

Lc0-v28 +53 elo

Code:

   PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)
   Stockfish 14           :  3685.9    9.9  3248.5    4100    79      99  2487  1523    90    37
   Komodo-Dragon 2        :  3587.4    8.6  3252.0    4500    72      67  2298  1908   294    42
   Lc0-v28                :  3584.2   12.9   591.5    1000    59     100   335   513   152    51
   Lc0-v27                :  3531.3   23.6   501.0     800    63      93   307   388   105    49

New results:
Lc0 now ranked 8th with 275/321

I read in talkchess that it even gets stronger when a special parameter is used:

Still 46 games to go.
First position in tournament has lc0 0.27.0 with 315,5/367.