Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Stockfish and Lc0: now and then
- By MrKris (***) Date 2020-12-26 04:14
Then:

Stockfish_20080616-1stNNUE   https://abrok.eu/stockfish/?page=8
Author: nodchip
Date: Thu Aug 6 16:37:45 2020 +0200  Timestamp: 1596724665
     Add NNUE evaluation {1st Sf-NNUE on abrok}
Net:  nn-97f742aaefcd_1stSf.nnue (aug 3)
https://tests.stockfishchess.org/nns?page=3

Lc0-then:  https://training.lczero.org/networks/1
64506 1 0e040a36 2951.00 2020-08-06 16:31:38 +00:00

Now:

Stockfish_20122320-now   https://abrok.eu/stockfish/
Author: Moez Jellouli {default embeded net}
Date: Wed Dec 23 20:20:24 2020 +0100  Timestamp: 1608751224
     Correct Outflanking calculations in classical eval

Lc0-now: net66792_dec23,2020 https://training.lczero.org/networks/1
66792  1 c20dfb7b 3058.00 2020-12-23 19:47:18 +00:00

All opeings: TCEC19_NoomenSelect.pgn

Four parts: SFnow - Lc0now, SFthen - Lc0then, SFthen - Lc0now, SFnow - Lc0then.

Part 1
G/3'+2'' 8GBh Sy.6 ; R7 2700X | RTX 2060, TCEC19_NoomenSelect.pgn
                             Elo   +-  Games  Score 
  1 Stockfish_20122320-now :  66   40   100   59.5%   +26 -7 =67
  2 Lc0v0.26.3_66792-now   :   0   37   100   40.5%  (draws 67.0%)

100 games white +30 -3 =67
Sf won both colors Rd. 26.  White won both games Rds. 6, 17, 28, 31 and 45 (5).
Games won that the opponent did NOT win the colors-reversed game:
  Stockfish: 21
        Lc0: 2 only!          {T19SuFi was Sf 54.5-45.5 Lc0}


Rd. 26, B81 Sicillian Scheveningen Keres
(game 51) Lc0 showed itself a modest disadvantage until it was totally lost by move 30: "30...Rcc7 {-0.55/13 14s} 31. Bxc5 {+5.00/32 2.7s} .  . ." [Event "G/3'+2'' 8GBh Sy.6"]
[Site "R7 2700X | RTX 2060"]
[Date "2020.12.25"]
[Round "26"]
[White "Stockfish_20122320-now"]
[Black "Lc0v0.26.3_66792-now"]
[Result "1-0"]
[ECO "B81"]
[Opening "Sicilian"]
[Variation "Scheveningen, Keres, 6...Nc6 7.g5 Nd7"]
[TimeControl "180+2"]
[Termination "adjudication"]
[PlyCount "96"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} c5 {book} 2. Nf3 {book} e6 {book} 3. d4 {book} cxd4 {book} 4.
Nxd4 {book} Nf6 {book} 5. Nc3 {book} d6 {book} 6. g4 {book} Nc6 {book} 7.
g5 {+1.44/30 0 8.1s} Nd7 {-0.69/9 0 2.7s} 8. h4 {+1.29/32 0 18s} Be7
{-0.62/10 0 5.2s} 9. Be3 {+1.21/31 0 13s} O-O {-0.60/10 0 2.6s} 10. a3
{+1.38/32 0 21s} Nb6 {-0.48/11 0 6.1s} 11. Qf3 {+1.51/28 0 3.4s} Ne5
{-0.49/13 0 6.7s} 12. Qg3 {+1.43/27 0 3.8s} Bd7 {-0.49/13 0 3.9s} 13. O-O-O
{+1.29/34 0 19s} Rc8 {-0.49/14 0 5.6s} 14. f4 {+1.34/29 0 1.9s} Nec4
{-0.47/15 0 3.6s} 15. Bf2 {+1.65/31 0 7.0s} d5 {-0.46/15 0 5.1s} 16. e5
{+1.40/30 0 3.5s} Qc7 {-0.57/11 0 11s} 17. Be1 {+1.38/29 0 3.1s} Na5
{-0.60/13 0 7.4s} 18. Rh2 {+1.32/31 0 13s} Nbc4 {-0.62/11 0 5.1s} 19. Na2
{+1.57/26 0 2.1s} Qb6 {-0.62/11 0 9.1s} 20. Bc3 {+1.58/29 0 3.4s} Nc6
{-0.57/13 0 3.6s} 21. Bxc4 {+1.44/28 0 3.0s} dxc4 {-0.56/12 0 2.1s} 22. h5
{+1.72/32 0 15s} Rfd8 {-0.51/14 0 5.8s} 23. g6 {+1.82/28 0 2.2s} fxg6
{-0.57/21 0 4.9s} 24. hxg6 {+2.83/33 0 13s} h6 {-0.59/20 0 2.8s} 25. Qg4
{+1.90/33 0 5.7s} Bc5 {-0.61/19 0 5.9s} 26. Nxc6 {+2.61/32 0 3.3s} Qxc6
{-0.62/22 0 4.4s} 27. Ba5 {+2.57/31 0 2.9s} b6 {-0.70/14 0 9.6s} 28. Bb4
{+2.98/29 0 3.7s} Be8 {-0.65/17 0 2.9s} 29. Rdh1 {+3.58/29 0 2.6s} Rd7
{-0.63/14 0 7.4s} 30. Nc3 {+3.84/30 0 2.7s} Rcc7 {-0.55/13 0 14s} 31. Bxc5
{+5.00/32 0 2.7s} bxc5 {-1.85/15 0 4.5s} 32. Rxh6 {+5.01/32 0 3.1s} gxh6
{-1.07/17 0 5.3s} 33. Rxh6 {+5.31/33 0 3.0s} Bxg6 {-1.36/18 0 5.2s} 34.
Qxg6+ {+5.30/32 0 3.0s} Rg7 {-2.75/18 0 3.9s} 35. Qh5 {+5.53/35 0 4.8s} Rh7
{-2.90/17 0 5.1s} 36. Qg5+ {+5.59/33 0 2.5s} Kf8 {-3.19/16 0 2.3s} 37. Qd8+
{+5.81/31 0 3.3s} Qe8 {-3.73/14 0 2.6s} 38. Qd6+ {+5.76/31 0 4.2s} Kg8
{-3.81/13 0 5.6s} 39. Rxe6 {+6.04/32 0 3.6s} Rce7 {-3.92/12 0 10s} 40. Rxe7
{+6.08/34 0 4.8s} Qxe7 {-4.13/12 0 2.0s} 41. Qd5+ {+6.09/32 0 3.5s} Qf7
{-4.33/10 0 6.6s} 42. Qe4 {+6.20/33 0 3.5s} Qg7 {-4.79/10 0 15s} 43. Kb1
{+7.65/25 0 3.4s} Qg1+ {-4.50/10 0 4.7s} 44. Ka2 {+8.65/27 0 3.9s} Qf1
{-4.42/8 0 4.8s} 45. Qg6+ {+70.33/35 0 2.5s} Rg7 {-5.70/11 0 7.1s} 46. Qe8+
{+M25/49 0 2.6s} Kh7 {-7.28/1 0 0s} 47. Qh5+ {+M23/82 0 2.5s} Kg8 {-10.01/1
0 0.003s} 48. Nd5 {+M21/88 0 2.6s} Rf7 {-17.42/8 0 12s, White wins by
adjudication} 1-0

Rd.26 (game 52) Lc0now 0-1 Sfnow: Same opening. Lc0's wasted move, 39.a4??, was enough to go from a draw to a loss in the sharp ending.
[FEN "3r3k/pr2qn1p/3p2p1/4p3/2p1P1Q1/P1P1B1R1/2P2P1P/2KR4 w - - 4 34"]
34. Rd5 {+0.33/13 7.5s} Rdb8 {0.00/38 2.7s} 35. h4 {+0.33/13 5.9s}
Rb1+ {0.00/43 4.3s} 36. Kd2 {+0.30/1 0.001s} Rh1 {0.00/45 16s}
37. Rh3 {+0.34/15 4.1s} Ra1 {0.00/43 2.5s} 38. h5 {+0.14/10 8.4s}
g5 {0.00/45 2.6s} 39. a4 {?? +0.17/11 5.2s} (39.Rg3 {! 0.00 ShashChess 1 min.} Rbb1 40.Ke2 Re1+ 41.Kf3 Qf6+ 42.Qf5 Qxf5+ 43.exf5 Kg7 44.Bxg5 {etc.})
39...Rbb1 {+1.56/30 2.7s}
40. Rb5 {-0.13/13 6.6s , relatively best, SC, but -3.22 at 2 mins., without 39.Rg3 white does not have enough counterplay}
40...Re1 {+3.06/31 6.2s} 41. h6 {-0.42/13 4.3s}
Qf8 {+3.78/30 3.4s} 42. a5 {-0.68/14 4.0s} a6 {+4.09/30 3.1s}
43. Rb6 {-1.60/14 6.3s} d5 {+4.39/32 3.4s} 44. exd5 {-1.91/15 6.6s}
Qd8 {+4.65/31 4.2s} 45. Rg3 {-2.09/16 7.4s} Qxd5+ {+4.85/29 4.2s}
46. Bd4 {-2.35/15 0.81s} Red1+ {+4.86/30 4.4s} 47. Qxd1 {-2.65/13 3.3s}
Rxd1+ {+5.00/27 2.9s} 48. Kxd1 {-2.79/12 2.1s} exd4 {+5.09/28 3.1s}
49. cxd4 {-2.91/11 5.1s} Kg8 {+5.18/27 3.7s} 50. c3 {-3.33/12 9.6s}
Nd6 {+5.55/25 3.1s} 51. Re3 {-3.49/12 3.4s} Kf7 {+6.07/28 3.5s}
52. Rxa6 {-3.96/11 5.4s} Nf5 {+6.77/24 2.3s} 53. Ree6 {-4.26/12 2.6s}
Qf3+ {+10.04/26 2.5s} 54. Kc2 {-4.84/12 5.0s} Qd3+ {+M35/60 2.2s}
55. Kb2 {-4.35/14 3.4s} Qd2+ {+M31/70 2.2s} 56. Kb1 {-4.85/14 2.2s}
Qd1+ {+M29/76 2.2s} 57. Kb2 {-6.74/14 6.4s} Qb3+ {+M27/80 2.4s}
58. Kc1 {-8.59/14 4.1s} Qxc3+ {+M25/82 2.8s} 59. Kb1 {-9.02/11 0.28s}
Nxd4 {+M21/87 2.1s} 60. Rf6+ {-11.32/10 3.1s, Black wins by adjudication 0-1}
Attachment: Sf-now_Lc-now.pgn - Games (274k)
- By MrKris (***) Date 2020-12-28 01:09
The first NNUE Stockfish on abrok.eu, Aug. 6 2020, vs. the official T60 net at the same date and time.

Part 2
G/3'+2'' 8GBh Sy.6 , R7 2700X | RTX 2060 , TCEC19_NoomenSelect.pgn
                                 Elo +- Games     Score  
  1 Stockfish_20080616-1stNNUE :  18 44  100  52.5%  +23 -18 =59
  2 lc0v0.26.3_64506-then      :   0 44  100  47.5%  (draws 59.0%)
                                                    white +41 -0 =59

                                                   
Rd.#'s = opening rev. game pair
Stockfish-1stN won 1.5/2: 6, 9, 16, 19, 24, 26, 27, 30, 34, 38, 41, 46, 47  (13)                            
  Lc0-T60-then won 1.5/2: 12, 13, 15, 22, 28, 20, 40, 43 (8)
          White won both: 2, 10, 17, 23, 25, 31, 32, 33, 39, 42 (10)
Opening wins: Stockfish 13 - 8 Lc0
         
Lc0 did not realize its king was too weak until too late: 19. Qxa7 {+3.35/29 3.9s} Bxf1 {-0.52/18 4.4s} 20. Rxf1 {+3.51/33 6.3s} Nf5 {-0.54/21 17s} 21. Rf3 {+3.70/33 3.2s} Nh4 {-2.65/13 13s}.
[Event "G/3'+2'' 8GBh Sy.6"]
[Site "R7 2700X | RTX 2060"]
[Date "2020.12.27"]
[Round "19"]
[White "Stockfish_20080616-1stNNUE"]
[Black "lc0v0.26.3_64506-then"]
[Result "1-0"]
[BlackElo "2400"]
[ECO "C18"]
[Opening "French"]
[Variation "Winawer, Poisoned Pawn, Konstantinopolsky Variation, 6.bxc3 Ne7 7.Qg4 Qc7 8.Qxg7 Rg8 9.Qxh7 cxd4 10.Ne2"]
[WhiteElo "2400"]
[TimeControl "180+2"]
[Termination "normal"]
[PlyCount "60"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} e6 {book} 2. d4 {book} d5 {book} 3. Nc3 {book} Bb4 {book} 4.
e5 {book} Ne7 {-0.21/11 0 4.2s} 5. a3 {+0.72/29 0 12s} Bxc3+ {-0.21/12 0
1.9s} 6. bxc3 {+0.72/25 0 2.0s} c5 {-0.19/11 0 2.9s} 7. Qg4 {+0.71/28 0
3.6s} cxd4 {-0.20/14 0 9.9s} 8. Qxg7 {+0.83/28 0 2.8s} Rg8 {-0.23/17 0
2.0s} 9. Qxh7 {+0.70/31 0 3.4s} Qc7 {-0.24/16 0 2.8s} 10. Ne2 {+0.76/30 0
4.4s} Nbc6 {-0.12/12 0 11s} 11. h4 {+0.74/31 0 21s} dxc3 {-0.14/11 0 4.1s}
12. f4 {+0.43/29 0 5.0s} Bd7 {-0.08/12 0 6.1s} 13. Qd3 {+0.29/31 0 10s}
O-O-O {-0.10/14 0 8.2s} 14. h5 {+0.44/31 0 7.7s} Nf5 {-0.08/17 0 1.7s} 15.
Rb1 {+0.74/26 0 2.3s} d4 {-0.12/19 0 4.3s} 16. Rg1 {+1.34/25 0 3.0s} Nh6
{-0.12/19 0 0.22s} 17. Nxd4 {+1.69/28 0 3.2s} Nxd4 {-0.37/16 0 6.6s} 18.
Qxd4 {+1.66/29 0 4.1s} Bb5 {-0.41/18 0 2.6s} 19. Qxa7 {+3.35/29 0 3.9s}
Bxf1 {-0.52/18 0 4.4s} 20. Rxf1 {+3.51/33 0 6.3s} Nf5 {-0.54/21 0 17s} (20.
.. Rxg2 21. Rf3 Rdg8 22. Qa8+ Kd7 23. Rd3+ Ke7 24. Rxb7 Rxa8 25. Rxc7+ Kf8
26. Rdxc3) 21. Rf3 {+3.70/33 0 3.2s} Nh4 {-2.65/13 0 13s} 22. Rh3 {+7.25/28
0 3.1s} Rxg2 {-1.88/11 0 3.7s} 23. Be3 {+8.33/28 0 3.8s} Rg4 {-2.75/11 0
6.5s} 24. Bb6 {+10.70/27 0 4.6s} Ng2+ {-3.65/14 0 3.5s} 25. Kf1 {+11.16/28
0 3.5s} Qc4+ {-4.30/15 0 5.1s} 26. Rd3 {+25.97/29 0 5.6s} Rxd3 {-4.61/17 0
3.6s} 27. Qa8+ {+50.31/32 0 3.0s} Kd7 {-5.04/1 0 0.001s} 28. Qd8+ {+M11/37
0 3.3s} Kc6 {-5.90/1 0 0.002s} 29. cxd3 {+M25/64 0 3.2s} Qxf4+ {-17.88/12 0
14s} 30. Bf2 {+M23/79 0 3.6s} Ne3+ {-19.57/12 0 2.9s, White wins by
adjudication} 1-0

Lc0 was quickly lost after 22...Kg6?, again underestimating its king's safety.
[Event "G/3'+2'' 8GBh Sy.6"]
[Site "R7 2700X | RTX 2060"]
[Date "2020.12.27"]
[Round "9"]
[White "Stockfish_20080616-1stNNUE"]
[Black "lc0v0.26.3_64506-then"]
[Result "1-0"]
[BlackElo "2400"]
[ECO "B45"]
[Opening "Sicilian"]
[Variation "Taimanov, 5.Nc3"]
[TimeControl "180+2"]
[Termination "normal"]
[PlyCount "74"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} c5 {book} 2. Nf3 {book} e6 {book} 3. d4 {book} cxd4 {book} 4.
Nxd4 {book} Nc6 {book} 5. Nc3 {book} d6 {book} 6. g4 {book} h6 {book} 7. h4
{book} a6 {book} 8. Be3 {+1.08/29 0 14s} Bd7 {-0.39/8 0 3.5s} 9. Be2
{+0.78/30 0 29s} Rc8 {-0.30/11 0 3.9s} 10. Nxc6 {+0.87/28 0 5.5s} Bxc6
{-0.34/13 0 4.7s} 11. Qd3 {+0.83/32 0 3.3s} Be7 {-0.25/13 0 5.8s} 12. Bd4
{+0.64/31 0 18s} e5 {-0.23/15 0 4.2s} 13. Be3 {+0.75/29 0 2.4s} Bxh4
{-0.21/14 0 3.2s} 14. O-O-O {+0.65/28 0 4.0s} Bg5 {-0.22/14 0 3.8s} 15. Qd2
{+0.91/33 0 11s} Bxe3 {-0.22/12 0 3.7s} 16. Qxe3 {+0.73/32 0 4.0s} g5
{-0.21/12 0 4.1s} 17. Rd2 {+0.67/27 0 3.7s} Qf6 {-0.21/12 0 8.6s} 18. Bc4
{+1.16/26 0 4.5s} Ne7 {-0.20/12 0 5.8s} 19. f3 {+0.85/29 0 6.2s} Rd8
{-0.20/11 0 5.6s} 20. a4 {+0.80/31 0 18s} Kf8 {-0.19/10 0 12s} 21. Kb1
{+0.96/28 0 2.1s} Kg7 {-0.28/10 0 9.1s} 22. Bb3 {+1.26/27 0 2.6s} Kg6
{-0.38/13 0 ? 8.2s a wasted move in an already disadvantaged position} (22.
.. Bd7 {! -1.07, +/, ShashChess 5 mins./45 plies}) 23. Ne2 {+2.13/28 0
3.2s} h5 {-0.38/12 0 2.6s} 24. gxh5+ {+2.27/28 0 3.3s} Rxh5 {-0.38/11 0
3.2s} 25. Rxh5 {+2.45/29 0 3.4s} Kxh5 {-0.41/16 0 6.5s} 26. Rd1 {+2.50/26 0
2.8s} Rg8 {-0.42/15 0 4.2s} 27. Rg1 {+2.65/29 0 3.2s} Kh6 {-0.83/12 0 12s}
28. Ng3 {+3.59/27 0 2.9s} Qg6 {-0.96/14 0 2.8s} 29. Rh1+ {+3.60/29 0 3.1s}
Kg7 {-1.21/13 0 6.6s} 30. Rh5 {+3.98/30 0 2.8s} Kf8 {-1.65/13 0 9.1s} 31.
Qb6 {+4.32/28 0 2.7s} Kg7 {-1.81/13 0 4.6s} 32. Qc7 {+4.21/35 0 14s} Qf6
{-1.83/15 0 6.3s} 33. Rh3 {+6.83/32 0 2.7s} Rh8 {-2.02/14 0 3.2s} 34. Rxh8
{+8.01/33 0 3.2s} Kxh8 {-6.07/13 0 6.5s} 35. Nh5 {+8.59/32 0 3.7s} Qxf3
{-7.82/13 0 8.9s} 36. Qxe7 {+8.97/31 0 3.4s} Qxh5 {-10.74/13 0 9.9s} 37.
Qf8+ {+9.30/30 0 2.7s} Kh7 {-18.01/1 0 0.001s, White wins by adjudication}
1-0
Attachment: Sf-then_Lc-then.pgn - Games (307k)
- By MrKris (***) Date 2020-12-28 20:28
The first NNUE Stockfish on abrok.eu (Aug. 6, 2020) vs. Lc0-recent net (Dec. 23, 2020).

Part 3
G/3'+2'' 8GBh Sy.6 , R7 2700X | RTX 2060 , TCEC19_NoomenSelect.pgn
                                 Elo +- Games     Score 
  1 Lc0v0.26.3_66792-now       :  60 45  100  58.5%  +30 =57 -13
  2 Stockfish_20080616-1stNNUE :   0 44  100  41.5%  (draws   57.0%)
                                                     white +36 -7 =57

                                                    
Rd.#'s = opening rev. game pair
  Sf-1stNNUE won: wh 14, 24, 26, 33, 38, 41; bk 4 =(7)
Lc0-T60-now won: wh 3, 7, 9, 12, 15, 17, 18, 25, 30, 31, 32, 34, 35, 48, 50 (15); bk 12, 36, 42 (3); both colors 23, 39, 46 (3x2) =(24)
     white won both: 6, 22, 27, 40, 45, 49 =(6)
Opening wins: Stockfish-1stN  7 - 24  Lc0-now

Lc0-now won with both colors after 1.e4 g6 {Book}.
Stockfish, white, lost after its evaluation declined to 0.00 for the first time, -2nd diagram-, black has its K on b8, its R(a8), B(c8), Q(d8) undeveloped and both its N's on the 7th rank.
[Event "G/3'+2'' 8GBh Sy.6"]
[Site "R7 2700X | RTX 2060"]
[Date "2020.12.28"]
[Round "23"]
[White "Stockfish_20080616-1stNNUE"]
[Black "Lc0v0.26.3_66792-now"]
[Result "0-1"]
[ECO "A42"]
[Opening "Modern"]
[Variation "c4 Pterodactyl"]
[TimeControl "180+2"]
[Termination "adjudication"]
[PlyCount "161"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} g6 {book} 2. d4 {+0.87/28 0 8.5s} Bg7 {-0.29/8 0 5.0s} 3. c4
{+0.97/25 0 1.7s} d6 {-0.26/9 0 4.1s} 4. Nc3 {+1.12/27 0 3.1s} c5 {-0.25/10
0 6.0s} 5. d5 {+1.33/23 0 2.3s} Bxc3+ {-0.24/10 0 3.1s} 6. bxc3 {+1.45/25 0
2.8s} e5 {-0.22/11 0 4.2s} 7. Bd3 {+1.33/30 0 13s} Nd7 {-0.14/8 0 5.2s} 8.
h4 {+1.54/29 0 5.3s} Qe7 {-0.08/8 0 8.5s} 9. f3 {+1.14/32 0 16s} Kd8
{-0.05/7 0 5.3s} 10. Ne2 {+1.53/26 0 7.5s} h5 {-0.03/7 0 5.0s} 11. Be3
{+1.59/31 0 8.5s} f6 {-0.03/8 0 8.8s} 12. Qd2 {+1.41/32 0 5.8s} Qf8
{-0.03/9 0 5.6s} 13. O-O {+1.49/27 0 6.5s} Nh6 {+0.04/9 0 2.7s} 14. Ng3
{+1.66/27 0 12s} Nf7 {+0.18/7 0 6.8s} 15. Rab1 {+1.26/28 0 19s} Kc7
{+0.15/7 0 5.8s} 16. Qc2 {+1.45/27 0 4.0s} Rg8 {+0.14/8 0 3.8s} 17. Qb3
{+1.02/28 0 9.1s} Qe8 {+0.08/7 0 8.9s} 18. a4 {+0.83/28 0 6.5s} a5 {+0.29/7
0 2.6s} 19. Qa2 {+0.04/26 0 20s} Qd8 {+0.27/7 0 8.6s}
20. Rb2 {0.0026 2.5s} Kb8 {+0.26/7 0 4.7s}
[FEN "rkbq2r1/1p1n1n2/3p1pp1/p1pPp2p/P1P1P2P/2PBBPN1/QR4P1/5RK1 w - - 4 21"]

{<<<<< A very strange position for black to win with !?} 21. Rbb1 {0.0026 18s} Nb6 {+0.31/6 0
5.7s} 22. Bf2 {0.0035 4.2s} Bd7 {+0.60/7 0 3.1s} 23. Rb3 {+0.40/33 0 6.3s}
Nc8 {+0.60/9 0 4.2s} 24. Rfb1 {-0.75/25 0 26s} b6 {+0.72/9 0 5.2s} 25. Qd2
{-0.55/25 0 9.8s} Bxa4 {+0.88/7 0 6.8s} 26. R3b2 {-0.39/24 0 2.6s} Bd7
{+0.83/7 0 7.0s} 27. Bc2 {-0.50/25 0 5.9s} Ra7 {+0.80/7 0 6.6s} 28. Qd1
{-0.64/28 0 4.4s} Qe8 {+0.81/7 0 3.1s} 29. Ra1 {0.0022 0.51s} Nh6 {+0.82/10
0 5.6s} 30. Qb1 {-0.66/24 0 3.7s} Kc7 {+1.17/9 0 4.2s} 31. Qc1 {-1.06/26 0
1.8s} g5 {+1.35/9 0 4.9s} 32. Nf1 {-1.27/25 0 2.4s} Qg6 {+1.30/10 0 4.3s}
33. Bb1 {-1.26/23 0 1.1s} Ne7 {+1.63/8 0 11s} 34. Bd3 {-2.16/23 0 3.0s} Nf7
{+1.67/9 0 4.4s} 35. Qd1 {-2.03/25 0 1.5s} Qg7 {+1.77/9 0 6.3s} 36. Qb3
{-2.32/24 0 1.8s} Rb8 {+2.44/9 0 3.3s} 37. hxg5 {-1.92/24 0 0.43s} Nxg5
{+3.18/8 0 5.0s} 38. Kh1 {-3.04/23 0 2.7s} h4 {+3.95/8 0 3.6s} 39. Qd1
{-2.73/20 0 0.31s} f5 {+4.89/8 0 4.1s} 40. exf5 {-2.87/23 0 1.4s} Nxf5
{+4.92/8 0 4.0s} 41. Bxf5 {-3.01/22 0 1.2s} Bxf5 {+4.73/8 0 4.2s} 42. Be3
{-2.86/23 0 1.9s} Raa8 {+4.20/8 0 13s} 43. Nd2 {-2.68/23 0 2.2s} Qg6
{+4.22/7 0 2.6s} 44. Qa4 {-2.25/26 0 0.87s} Bd7 {+4.40/8 0 3.1s} 45. Qc2
{-2.46/26 0 0.78s} Qxc2 {+6.57/8 0 4.1s} 46. Rxc2 {-2.58/27 0 1.4s} Bf5
{+6.54/8 0 8.5s} 47. Rb2 {-2.58/28 0 2.1s} Rg8 {+6.60/7 0 1.6s} 48. Re1
{-2.36/29 0 0.61s} Bd3 {+5.47/7 0 7.2s} 49. Bxg5 {-2.40/33 0 1.2s} Rxg5
{+5.71/8 0 1.9s} 50. Re3 {-2.40/41 0 1.1s} Bf5 {+5.87/7 0 2.9s} 51. Ne4
{-2.40/41 0 6.6s} Rgg8 {+5.45/7 0 7.3s} 52. Kh2 {-2.40/35 0 0.88s} a4
{+5.82/7 0 2.7s} 53. Ra2 {-2.58/32 0 3.2s} a3 {+6.89/7 0 2.1s} 54. Nd2
{-2.85/32 0 3.0s} Ra4 {+7.11/7 0 4.2s} 55. Re1 {-2.89/30 0 2.3s} Bd3
{+5.84/7 0 3.3s} 56. Re3 {-2.85/29 0 0.90s} Bg6 {+5.50/8 0 4.4s} 57. Kg1
{-2.66/31 0 3.5s} Rb8 {+6.74/6 0 2.9s} 58. Re1 {-2.85/30 0 0.56s} Bd3
{+6.03/7 0 2.4s} 59. Re3 {-3.69/34 0 6.6s} Bxc4 {+7.20/9 0 2.3s} 60. Nxc4
{-3.66/25 0 0.46s} Rxc4 {+6.91/9 0 1.00s} 61. Rxa3 {-4.26/28 0 1.4s} b5
{+6.86/9 0 1.8s} 62. Ra7+ {-4.06/27 0 1.5s} Rb7 {+7.69/9 0 2.2s} 63. Ra6
{-4.18/29 0 1.4s} b4 {+7.52/9 0 3.0s} 64. Rc6+ {-4.46/25 0 3.2s} Kd7
{+7.41/9 0 1.7s} 65. cxb4 {-4.56/23 0 0.49s} Rbxb4 {+8.51/9 0 4.7s} 66. Kh2
{-4.57/26 0 4.1s} Rd4 {+18.45/7 0 1.7s} 67. Ra6 {-4.52/22 0 0.31s} Rxd5
{+26.15/6 0 2.2s} 68. Ra7+ {-4.84/24 0 2.8s} Ke6 {+26.01/6 0 4.3s} 69. Rc3
{-4.81/25 0 1.6s} Rd2 {+31.74/5 0 2.4s} 70. Rc1 {-4.81/25 0 2.0s} Rbb2
{+24.26/5 0 2.2s} 71. Rg7 {-5.65/28 0 2.8s} Kd5 {+23.30/5 0 2.0s} 72. Kh3
{-6.17/27 0 4.5s} c4 {+25.32/5 0 2.1s} 73. Rg8 {-6.33/23 0 1.3s} Rbc2
{+18.55/5 0 2.1s} 74. Ra1 {-6.60/21 0 1.6s} c3 {+30.45/5 0 2.1s} 75. Ra5+
{-7.48/22 0 3.4s} Kd4 {+21.59/5 0 2.1s} 76. Rg4+ {-6.79/19 0 0.38s} Ke3
{+39.34/6 0 1.7s} 77. Ra1 {-8.15/21 0 3.6s} d5 {+51.31/5 0 2.5s} 78. Rg5
{-7.51/18 0 0.63s} e4 {+44.84/5 0 2.2s} 79. fxe4 {-7.36/21 0 1.2s} dxe4
{+39.30/5 0 1.9s} 80. Rc5 {-9.29/22 0 4.1s} Rd3 {+41.72/5 0 2.1s} 81. Rf5
{-10.40/19 0 2.4s, Black wins by adjudication 0-1}
Attachment: Sf-then_Lc-now.pgn - Games (286k)
- By MrKris (***) Date 2020-12-31 00:01
The Lc0 from then (8-6-2020=Sf's 1st NNUE) vs. Stockfish now (12-23-2020).

Part 4
G/3'+2'' 8GBh Sy.6 , R7 2700X | RTX 2060 , TCEC19_NoomenSelect.pgn
                               Elo  +- Games     Score 
  1 Stockfish_20122320-now  :  134  44  100  68.5%  +39 -2 =59
  2 lc0v0.26.3_64506-then   :    0  39  100  31.5%  (draws 59.0%)
 
Stocfish won 39 and lost only 2 vs. the 'then' Lc0 net.
But both of the 2 were 'white-won both games' openings,
so it was 37 - 0 opening wins for Stockfish!
 
Parts 1 through 4, % per match and Elo difference:

            totals/200                            Lc0now/100   Lc0then/100
Stockfish-now  * 128.0                            59.5% (+66)  68.5% (+134) **
Sf-1stNNUEabrok * 94.0                            41.5% (-60)  52.5% (+18)
     Sf Team=222.0/400
                         Sfnow/100    Sf1stN/100
Lc0 T60-now      @ 99.0   40.5% (-66)  58.5% (+60) ^^
Lc0 T60-then     @ 79.0   31.5% (-134) 47.5% (-18)
      Lc0 Team=178.0/400


Since the 1st NNUE Stockfish has improved more than Lc0:

* Stockfish from 94.0/200 to 128.0/200 vs. the now and then Lc0's,
@ Lc0 from 79.0/200 to 99.0/200 vs. the now and then Stockfish's.

** From Stockfish-now's view Lc0 increased 68 Elo then to now,
^^ From Lc0-now's view Stockfish increased 126 Elo then to now.

 
Stockfish was >5.00 winning at move 24, and its strong attack showed why.
[Event "G/3'+2'' 8GBh Sy.6"]
[Site "R7 2700X | RTX 2060"]
[Date "2020.12.30"]
[Round "23"]
[White "Stockfish_20122320-now"]
[Black "lc0v0.26.3_64506-then"]
[Result "1-0"]
[BlackElo "2400"]
[ECO "B06"]
[Opening "Modern"]
[Variation "3.Nc3 d6 4.Be3 a6"]
[WhiteElo "2400"]
[TimeControl "180+2"]
[Termination "normal"]
[PlyCount "80"]
[WhiteType "human"]
[BlackType "human"]

1. e4 {book} g6 {book} 2. d4 {+0.89/28 0 6.0s} d6 {-0.27/7 0 4.6s} 3. Nc3
{+0.88/28 0 4.6s} Bg7 {-0.27/8 0 4.2s} 4. Be3 {+0.73/28 0 4.3s} a6 {-0.28/8
0 5.1s} 5. Nf3 {+0.84/31 0 16s} Nf6 {-0.25/9 0 3.8s} 6. Qd2 {+0.79/31 0
22s} b5 {-0.21/12 0 8.2s} 7. Bh6 {+0.49/30 0 8.6s} O-O {-0.18/13 0 2.6s} 8.
Bd3 {+0.50/31 0 5.0s} Nc6 {-0.20/15 0 2.8s} 9. Nd5 {+0.62/30 0 2.7s} Re8
{-0.18/13 0 12s} 10. Bxg7 {+0.67/29 0 5.5s} Kxg7 {-0.17/12 0 2.4s} 11. h4
{+0.71/29 0 3.3s} h5 {-0.16/14 0 3.5s} 12. Ne3 {+0.70/30 0 6.3s} e5
{-0.17/17 0 3.8s} 13. d5 {+0.42/31 0 5.2s} Ne7 {-0.20/15 0 3.8s} 14. a4
{+0.32/32 0 16s} Bd7 {-0.07/11 0 7.2s} 15. O-O {+0.21/33 0 18s} bxa4
{-0.06/12 0 6.4s} 16. Ng5 {+0.12/28 0 1.8s} Neg8 {-0.08/12 0 10.0s} 17. f4
{+0.12/32 0 7.3s} Qe7 {-0.08/14 0 7.1s} 18. Rf3 {+0.65/29 0 5.6s} Rab8
{-0.03/10 0 3.3s} 19. Nc4 {0.0036 11s} exf4 {+0.21/14 0 5.7s} 20. Rxf4
{+0.94/30 0 2.1s} Bb5 {+0.24/16 0 2.4s} 21. Na5 {+0.50/37 0 5.1s} Qe5
{+0.10/13 0 ? 5.5s +/ or +-} (21. .. Bd7 {! Stockfish 1 min., -0.35/35,
+=}) 22. Bxb5 {+2.37/26 0 2.3s} Qxb2 {-0.34/14 0 ? 5.7s +-} (22. .. Rxb5
{!} 23. Nc4 Qe7 24. Qd4 {with a continuing attack}) 23. Raf1 {+4.63/28 0
2.6s} axb5 {-0.59/15 0 9.9s} 24. Nc6 {+5.93/27 0 2.4s} a3 {-0.59/10 0
0.21s} 25. Qf2 {+6.51/30 0 2.4s} a2 {-0.83/19 0 6.1s} 26. e5 {+7.48/32 0
3.2s} a1=Q {-1.33/22 0 13s}
[FEN "1r2r1n1/2p2pk1/2Np1np1/1p1PP1Np/5R1P/8/1qP2QP1/q4RK1 w - - 0 27"]

27. Rxf6 {+8.73/32 0 2.4s --Stockfish ignore's black's 2nd Q !--} Nxf6 {-1.39/23 0 0.28s --Lc0's old net's evaluation is 'clueless'.--} 28. Qxf6+ {+10.20/29 0 3.4s} Kh6
{-1.47/20 0 0.20s} 29. Nxf7+ {+10.78/29 0 2.9s} Kh7 {-0.67/1 0 0s} 30. Ng5+
{+11.48/34 0 11s} Kh6 {-2.72/16 0 9.3s} 31. Nf7+ {+12.92/35 0 11s} Kh7
{-0.68/1 0 0s} 32. Rxa1 {+14.21/34 0 10s} Qxa1+ {-2.92/17 0 11s} 33. Kh2
{+14.50/34 0 2.6s} Ra8 {-2.96/17 0 8.6s --white is winning 2 exchanges
down--} 34. Ng5+ {+14.93/35 0 3.0s} Kh6 {-3.19/18 0 0.29s} 35. Qf7
{+16.09/35 0 4.8s} Rh8 {-2.87/19 0 9.0s} 36. Ne6 {+16.41/33 0 2.4s} Rhg8
{-3.84/15 0 5.3s} 37. Qf4+ {+16.73/31 0 2.1s} Kh7 {-4.61/14 0 1.7s} 38.
Ng5+ {+17.20/31 0 2.1s} Kh6 {-8.96/12 0 19s} 39. Nd4 {+19.34/33 0 12s} Rg7
{-9.81/10 0 3.4s} 40. e6 {+M21/65 0 1.0s} Re7 {-10.11/9 0 6.1s, White wins
by adjudication 1-0} 1-0
Up Topic The Rybka Lounge / Computer Chess / Stockfish and Lc0: now and then

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill