Stockfish demonstrates insane level of playing strength.

In this test.

I played 50 games of chess at 1 second per game, or just microsecond of thinking time per move.

The games were played between Stockfish 15 DEV vs Dragon 3.1. They are the strongest chess engines on the planet.

banksia game  2022

                                 
1   Stockfish dev-20221221  +346  +38/=12/-0 88.00%   44.0/50
2   Dragon 3.1              -346  +0/=12/-38 12.00%    6.0/50


It is not shocking that Stockfish won this micro bullet match.

What is shocking is the level of play Stockfish demonstrated at this micro bullet match.

Here are the centipawn analysis results from this match.

Dragon 3.1:   38/4/22/14/2/34/62/70/133/28/43/24/40/33/46/5/46/25/79/41/85/66/30/1/27/76/6/38/29/37/32/22/37/29/38/25/19/4/6/29/50/38/44/29/22/61/5/31/2/0  => Average=0.35

Dragon played at a grandmaster level of play for a grandmaster playing a chess game at 3 minutes a move, or at standard FIDE time controls.
When playing chess at 1 second per game. Dragon 3.1 had a average move error rate of .35 Centipawns per move.

As a frame of reference I took the best games of GM Magnus Carlsen. And the games were all wins, and played in the last 7 years.
So this is not GM Carlsen's average error rate, but the best error rate of his best cherry picked games.

GM Magnus Carlsen's best error rate is .12 Centipawns per move. In his best 50 games. Played in the last 7 years. And that were all wins at 3 minutes per move. Or FIDE standard time controls.

Carlsen Magnus:   2/10/10/21/14/23/23/4/30/22/49/8/4/15/18/35/6/8/16/5/11/4/9/13/14/7/5/12/7/20/12/27/7/14/1/13/8/2/18/7/11/10/10/6/4/10/2/4/13/7  => Average=0.12

Stockfish at 1 second per game scored an incredible .07 centipawn loss per move average in 50 games.

Stockfish dev-20221221:   7/4/10/14/1/4/2/2/41/6/3/12/1/17/2/4/3/4/17/5/0/11/1/1/3/26/6/2/2/10/14/2/4/28/9/3/2/3/5/8/2/4/6/5/9/8/4/7/2/0  => Average=0.07

Stockfish demonstrated a level of play much better then GM Magnus Carlsen can demonstrate in his best games. On his best days, from the last 7 years. Played at 3 minutes a move, or FIDE standard time controls.

At 1 Second Per Game.

Did you check for time forfeits?

Admin wrote:

Did you check for time forfeits?

No time forfeit.

And I was able to watch all 50 games. As it took about 2 minute to play the match of 50 games at 1 second a game time controls.

Then I did the analysis of all 50 games and that took a few hours.

And a time forfeit would not affect the centipawn scoring error rate of the games. A error rate of just .07 pawns is just amazing when the chess engine only has
microseconds of processing time per move. To complete the game without time forfeit.

It is not so shocking how much better Stockfish played then Dragon. As you can see that in bullet as well, but not to that extent.

What shocked me was at 1 second a game. A chess engine could play better then any human chess player in the history of chess at tournament time controls.

Now it is very easy to understand why you do not need massive hardware to have great analysis with Stockfish 15 DEV.

I will post some games.

Code:

------------------------------------------------------------------------------------
  #  name                    games    wins   draws  losses   score    elo    +    -
  1. Stockfish dev-20221221     50      38      12       0      44   3938   53   45
  2. Dragon 3.1                 50       0      12      38       6   3662   45   53

Cross table:
------------------------------------------------------------------------------------
  #  name                       score   games         1         2
  1. Stockfish dev-20221221        44      50         x      44.0
  2. Dragon 3.1                     6      50       6.0         x

Tech:
------------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  #  name                      nodes/m         NPS  depth/m   time/m    moves     time
  1. Stockfish dev-20221221       354K    15453166     14.3      0.0     61.7      1.4
  2. Dragon 3.1                   113K    10671436      9.4      0.0     61.6      0.7
     all ---                      228K    13940597     11.8      0.0     61.6      1.0

These are the best 5 games of the match. As judged by Chessbase 17's Beauty Algorithm!
Included is the analysis.


[pgn] [Event "banksia game"]
[Site "*"]
[Date "2022.12.31"]
[Round "1"]
[Board "45"]
[White "Stockfish dev-20221221"]
[Black "Dragon 3.1"]
[Result "1-0"]
[ECO "C50"]
[PlyCount "158"]
[Beauty "5365487958608"]
[EventDate "2022.??.??"]
[TimeControl "1"]

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. d3 Bc5 5. c3 {C50: Giuoco Pianissimo} d6 6. O-O O-O {End of opening. Stockfish dev-20221224-64656f85:} 7. h3 {0.27/16 89 1289503 28/970/2} h6 {0.51/7 10 53948} 8. Re1 {0.28/14 30 489087 29/970/1} a5 {0.56/9 12 193150} 9. Bb3 {0.34/16 88 1311071 38/961/1} Be6 {0.42/9 13 180519} 10. Ba4 {0.39/12 10 155417 46/953/1. OK} Ba7 {0.62/9 10 131235. OK} 11. d4 {0.41/13 23 379652 51/948/1} Bd7 {0.35/10 11 187889. Best} 12. Bc2 {0.23/16 142 2304128 20/978/2. OK} Re8 {0.44/9 10 99544. OK} 13. a4 {0.16/14 33 520061 13/985/2. OK} Bb6 {0.43/9 13 219896} 14. Be3 {0.32/13 15 233658 31/968/1. OK} exd4 {0.24/9 10 78140. Best} 15. Bxd4 {0.34/15 28 451919 33/966/1} Nxd4 {0.35/9 10 108454} 16. cxd4 {0.36/11 10 127282 35/964/1} Nh7 {0.37/10 12 206910. OK} 17. Nc3 {0.38/18 87 1254568 39/960/1} Ng5 {0.28/9 10 129296} 18. Nxg5 {0.63/13 13 201649 133/867/0. OK} Qxg5 {0.27/9 10 125250} 19. Qd3 {0.64/14 30 497643 138/862/0. OK} c6 {0.41/8 10 98711} 20. Rad1 {0.54/13 15 244580 86/914/0} g6 {0.55/9 12 196309} 21. Re3 {0.62/12 20 327192 125/875/0. OK} h5 {0.61/9 12 192308. OK} 22. Ne2 {0.67/13 22 362044 157/843/0. OK} Rab8 {0.84/9 10 113062} 23. f4 {0.73/13 17 269493 200/800/0} Qh6 {0.90/9 10 74283. OK} 24. Kh1 {0.81/13 40 616156 283/717/0} h4 {0.89/8 10 123745} 25. Rf3 {0.88/14 20 302306 357/643/0} Bd8 {0.81/9 12 195419. OK} 26. e5 {0.93/10 10 129730 427/573/0} Be6 {0.71/9 10 116749. OK} 27. Nc3 {0.84/15 111 1683497 306/694/0} Be7 {0.96/9 11 189469. OK} 28. Ne4 {0.92/12 12 218053 408/592/0. OK} Rbd8 {0.99/10 13 235489. Inaccurate} 29. Rdf1 {1.00/13 18 262100 503/497/0. Mistake (0.7)} (29. exd6 $16) 29... Bf5 $14 {0.97/9 10 164559. OK} 30. Qc3 {1.07/12 27 436553 598/402/0} Qg7 {1.00/8 10 81425. OK} 31. Bd3 {1.15/15 15 256781 696/304/0} c5 {1.20/8 12 194421. Mistake (0.} (31... Bxe4 $14) 32. dxc5 {1.30/11 10 101048 839/161/0. Inaccurate} dxc5 {1.30/8 10 100015} 33. Nf2 {1.23/11 13 214044 779/221/0} Rd4 $1 {1.79/9 10 170328. Strong} 34. Ng4 {0.81/17 76 1258949 261/739/0. OK} Red8 $1 {1.33/10 10 153683. Strong} 35. Bc4 {0.80/12 13 222021 249/751/0. Inaccurate} Kh8 {1.00/9 12 181759} 36. b3 {0.84/10 13 201361 292/708/0} b6 {1.26/9 10 126619. OK} 37. R3f2 {0.88/11 10 125441 335/665/0. OK} Bc8 {1.05/8 10 165394. Mistake (0.} (37... f6 $14) 38. Qe1 {1.01/12 26 392320 502/498/0. Inaccurate} Bf5 {1.29/9 14 243378. OK} 39. Ne3 {0.99/12 22 345639 474/526/0} Qh7 {1.23/9 13 201803. Inaccurate} 40. Qe2 {1.24/13 13 205116 772/228/0} Kg7 {1.38/9 10 [#] 109279. Mistake (0.9)} (40... Bf8 $16) 41. e6 $1 $18 {1.52/14 10 145514 938/62/0. Strong} f6 {1.32/9 17 258711. OK} 42. Nxf5+ {1.80/10 10 81297 985/15/0. OK} gxf5 {1.40/9 10 115616} 43. Qf3 {1.62/15 44 731043 962/38/0. OK} Kh6 $2 {1.15/8 10 56478. Mistake (1.6)} (43... Kh8) 44. Rg1 $18 {1.71/12 29 471833 974/26/0. Best} Rg8 {1.26/10 10 95933} 45. Qc6 $2 {1.17/14 28 492229 691/309/0. Missed win (3.0)} (45. g3 $18) 45... Rd6 $16 {1.10/8 10 87581. Best} 46. Qf3 {1.28/9 10 142445 793/207/0} Rg3 {1.01/8 10 87616. Best} 47. Qe2 {0.87/12 27 446244 309/691/0. OK} Rd4 {0.61/8 10 121009. OK} 48. Qc2 {0.84/12 11 175007 278/722/0. Inaccurate} Qg8 {0.70/9 12 193771. Inaccurate} 49. Rf3 {0.82/11 10 88564 254/746/0. OK} Qg7 {0.64/8 10 84149. Inaccurate} 50. Bd3 {0.89/11 20 311191 327/673/0} Bd6 {0.64/8 10 118047. Mistake (1.1)} (50... Qg6 $16) 51. Bxf5 $18 {1.84/10 10 52168 985/15/0. OK} Bxf4 $2 {0.90/9 10 109009. Mistake (3.9)} (51... Be7) 52. Rxf4 $18 {3.75/12 10 41435 1000/0/0} Rxf4 {3.97/8 10 104483} 53. Qd2 {3.97/11 10 69670 1000/0/0} Kg5 {6.30/10 11 134029} 54. Rf1 {4.44/11 10 100664 1000/0/0} Qc7 {7.12/10 18 300852. OK} 55. e7 {3.97/13 20 124491 1000/0/0} Kxf5 {7.41/10 10 94813} 56. Rxf4+ {3.91/9 10 101695 1000/0/0} Qxf4 {7.36/8 10 96884} 57. Qd5+ {3.93/10 10 56483 1000/0/0} Qe5 {7.52/9 10 68852} 58. Qxe5+ {4.01/10 10 65067 1000/0/0} fxe5 {7.28/10 10 133389} 59. e8=Q {4.03/10 10 104927 1000/0/0} Rc3 {7.09/9 10 70174} 60. Qh5+ {4.21/10 13 200540 1000/0/0} Ke4 {7.41/9 10 57308} 61. Qxh4+ {4.24/10 10 60323 1000/0/0. Cut Off at move 61} Kd3 {8.16/8 10 113323} 62. Qg3+ {4.36/8 10 71576 1000/0/0} Kc2 {6.69/8 10 94962} 63. Qxe5 {4.50/9 10 44794 1000/0/0} Rxb3 {7.51/8 10 64079} 64. g4 {4.78/9 10 57330 1000/0/0} Rb1+ {9.12/8 13 182123} 65. Kg2 {5.54/9 10 109016 1000/0/0} Rb2 {8.86/8 10 59298} 66. g5 {6.16/8 10 68052 1000/0/0} Kb1+ {10.82/8 10 68396} 67. Kg1 {6.37/7 10 40329 1000/0/0} Rb4 {10.54/8 10 91596} 68. g6 {6.87/8 10 43802 1000/0/0} Rd4 {13.79/8 10 59296} 69. g7 {7.02/7 10 84971 1000/0/0} Rd1+ {13.68/8 10 84176} 70. Kg2 {7.40/9 10 117565 1000/0/0} Rd2+ {13.46/8 10 88643} 71. Kf3 {7.67/10 10 128981 1000/0/0} Rd8 {14.94/8 10 57544} 72. h4 {8.15/8 10 86104 1000/0/0} Kc2 {15.32/7 10 106878} 73. h5 {8.83/8 10 80343 1000/0/0} Kb3 {16.72/9 11 166785} 74. h6 {10.03/9 10 161457 1000/0/0} Kb4 {21.29/9 10 91352} 75. h7 {11.40/11 10 124308 1000/0/0} Rd3+ {26.81/8 10 43219} 76. Kf4 {13.14/11 10 185029 1000/0/0} Kxa4 {31.23/9 10 120967} 77. g8=Q {87.94/12 10 58732 1000/0/0} Rd4+ {10 M-7/9 84892} 78. Qxd4+ {87.94/16 10 16539 1000/0/0} cxd4 {10 M-6/3 100} 79. Qb3+ {87.95/10 10 9274 1000/0/0} Kxb3 {10 adjudication by egtb +12.71/5 4208.  Loses game:     ---       ---        Missed win:     White=1     ---        Mistake:       White=1     Black=6      Inaccurate:     White=4     Black=4      OK:         White=16     Black=15     Best:         White=1     Black=4      Strong:       White=1     Black=2      Brilliant:       ---       ---       .  Weighted Error Value: White=0.17/Black=0.33 Centipawn loss: w=14/b=33} 1-0

[/pgn]

[pgn][Event "banksia game"]
[Site "*"]
[Date "2022.12.31"]
[Round "1"]
[Board "26"]
[White "Dragon 3.1"]
[Black "Stockfish dev-20221221"]
[Result "0-1"]
[ECO "E12"]
[PlyCount "119"]
[Beauty "6393058876986"]
[EventDate "2022.??.??"]
[TimeControl "1"]

1. d4 Nf6 2. c4 b6 3. Nf3 e6 4. Nc3 {Stockfish dev-20221224-64656f85:} Bb7 {E12: Queen's Indian, 4.Nc3} 5. Bf4 Be7 {End of opening} 6. Qc2 {0.79/10 14 168279} c5 {0.20/17 115 1840475 2/979/19} 7. e4 {0.58/10 15 241895. OK} cxd4 {0.21/14 12 219092 2/978/20. OK} 8. Nxd4 {0.50/9 10 116128} d6 {0.22/14 12 207610 2/977/21. OK} 9. Bg3 {0.64/10 10 146046} Nh5 {0.18/14 15 244217 2/982/16. OK} 10. O-O-O {0.49/8 10 82430. OK} Nxg3 {0.25/18 89 1439654 2/976/22. OK} 11. hxg3 {0.62/10 16 242256} Nc6 {0.22/15 11 231169 2/979/19. OK} 12. e5 {0.44/11 13 222225} Nxe5 {0.16/15 33 541808 2/984/14. OK} 13. Rxh7 {0.87/10 10 168234. OK} Rxh7 {0.15/13 13 221070 2/985/13} 14. Qxh7 {0.98/9 10 93386} Bf8 {0.15/14 19 319889 2/986/12. OK} 15. f4 {1.22/10 10 126569. OK} Nc6 {0.17/15 25 461915 2/985/13. OK} 16. Ndb5 {0.21/9 10 101280. Inaccurate} a6 {-0.43/14 26 422283 50/950/0. Best (Easy)} 17. Nxd6+ $2 {0.00/9 10 98085. Loses game (1.9)} (17. Na3 $11) 17... Bxd6 $19 {-1.37/14 20 287307 897/103/0} 18. Ne4 {-0.60/9 10 190972} Nd4 $1 {-1.18/16 46 663364 747/253/0. Strong} 19. Qxg7 {-1.52/10 10 [#] 184748. OK} Nf5 {-1.47/15 37 592179 936/64/0. OK} 20. Nxd6+ {-1.23/10 10 93124. OK} Nxd6 {-1.58/12 16 281371 963/37/0} 21. Qe5 {-1.21/9 18 348232. OK} Ke7 {-1.58/15 31 521166 963/37/0. OK} 22. f5 {-1.49/9 10 130879. OK} Qc7 {-1.63/13 44 747283 972/28/0. Best (Easy)} 23. b4 $2 {-1.50/9 10 [#] 82679. Mistake (2.4)} (23. fxe6) 23... Rc8 $1 $19 {-1.83/12 11 183077 990/10/0. Strong} 24. fxe6 {-1.93/10 15 221196} Nxc4 {-4.65/16 82 1366417 1000/0/0. Best} 25. Qg5+ {-0.74/11 10 63759} Kxe6 {-4.77/13 25 394968 1000/0/0. Best (Easy)} 26. Qh6+ {-3.31/10 10 158781. OK} f6 {-4.65/18 56 934359 1000/0/0. OK} 27. Qh3+ {-4.02/9 10 43166. OK} Ke5 {-4.61/15 13 222601 1000/0/0. OK} 28. Bd3 {-3.60/8 10 73532} Be4 {-4.68/12 24 390588 1000/0/0. OK} 29. Bxe4 {-3.56/7 10 64890. OK} Nd6+ {-4.73/12 39 670854 1000/0/0} 30. Kb2 {-5.54/10 10 74671} Nxe4 {-4.75/11 10 195662 1000/0/0} 31. Qd7 {-5.85/9 10 47133} Qxd7 {-4.72/13 11 200238 1000/0/0} 32. Rxd7 {-5.69/8 10 79990} Rc6 {-4.76/11 11 180606 1000/0/0} 33. Rg7 {-5.53/7 10 97173} b5 {-4.84/10 18 319159 1000/0/0. OK} 34. a3 {-5.95/8 10 88755} Rd6 {-4.95/15 104 1651524 1000/0/0. Black is clearly winning} 35. Re7+ {-6.20/8 10 54883} Re6 {-4.99/11 14 214713 1000/0/0} 36. Rg7 {-6.20/8 10 105531} Kd4 {-4.94/14 54 893163 1000/0/0} 37. Kb3 {-6.18/8 10 69814} Nd2+ {-5.22/11 10 136891 1000/0/0} 38. Kc2 {-6.52/8 10 47793} Nc4 {-5.37/10 10 130366 1000/0/0} 39. Rd7+ {-6.54/7 10 63035} Rd6 {-5.42/11 10 112444 1000/0/0} 40. Rc7 {-6.61/8 10 106240} Nxa3+ {-5.81/12 18 258523 1000/0/0} 41. Kb3 {-6.92/7 10 49452} Nc4 {-5.85/10 13 181423 1000/0/0} 42. Kc2 {-7.42/8 10 106255} Ne3+ {-6.08/10 44 640611 1000/0/0} 43. Kd2 {-7.35/8 10 88354} Nd5 {-6.19/11 43 581244 1000/0/0} 44. Rh7 {-7.39/8 10 95706} Nxb4 {-6.36/7 10 118702 1000/0/0} 45. Ra7 {-8.08/8 10 132651} Kc4+ {-6.99/9 12 162153 1000/0/0} 46. Ke3 {-8.80/8 10 125262} Nd5+ {-7.05/12 15 215095 1000/0/0} 47. Kf3 {-8.98/8 10 53682} b4 {-7.35/11 10 132312 1000/0/0} 48. Rb7 {-11.00/9 10 128175} b3 {-7.71/11 15 262294 1000/0/0} 49. Rb8 {-11.61/9 10 122152} Nb4 {-8.22/9 17 253695 1000/0/0} 50. Rh8 {-11.46/8 10 45486} Rd3+ {-8.51/9 25 357814 1000/0/0} (50... b2 $142) 51. Kg4 {-11.40/7 10 35268} b2 {-9.14/9 10 86559 1000/0/0} 52. Rh1 {-12.76/8 10 61216} a5 {-9.71/9 10 126904 1000/0/0} 53. Kf5 {-13.86/8 12 197689} a4 {-9.82/10 23 306887 1000/0/0} 54. Kg6 {-15.42/8 10 96885} Rxg3+ {-11.28/12 24 386958 1000/0/0} 55. Kh6 {-13.98/8 10 57531} a3 {-12.66/9 10 132196 1000/0/0} 56. Rh4+ {-15.57/9 10 95514} Kb3 {-18.72/12 10 59719 1000/0/0} 57. Rf4 {-22.84/9 10 122167} b1=Q {-87.90/11 10 38705 1000/0/0} 58. Rf3+ {-22.14/9 10 82507} Rxf3 {-87.91/14 10 23157 1000/0/0} 59. gxf3 {-25.74/11 10 204990} Qg6+ {-87.95/20 10 52963 1000/0/0} 60. Kxg6 {10 adjudication by egtb -15.17/5 5124.  Loses game:     White=1     ---        Missed win:     ---       ---        Mistake:       White=1     ---        Inaccurate:     White=1     ---        OK:         White=11     Black=14     Best:         ---       Black=4      Strong:       ---       Black=2      Brilliant:       ---       ---       .  Weighted Error Value: White=0.46/Black=0.02 Centipawn loss: w=32/b=3} 0-1

[/pgn]

[pgn][Event "banksia game"]
[Site "*"]
[Date "2022.12.31"]
[Round "1"]
[Board "11"]
[White "Stockfish dev-20221221"]
[Black "Dragon 3.1"]
[Result "1-0"]
[ECO "B84"]
[PlyCount "45"]
[Beauty "8248509619255"]
[EventDate "2022.??.??"]
[TimeControl "1"]

1. e4 c5 2. Ne2 d6 3. Nbc3 a6 4. d4 cxd4 5. Nxd4 e6 6. Be2 Nf6 {End of opening} 7. f4 {0.49/16 135 2052521 83/916/1; B84: Sicilian, Scheveningen (Paulsen), classical variation} Be7 {0.55/8 10 116736. Stockfish dev-20221224-64656f85:} 8. Be3 {0.60/10 10 148263 136/864/0} O-O {0.60/9 12 159452} 9. g4 {0.72/11 14 221896 216/784/0} d5 {0.72/11 12 193372} 10. e5 {0.65/11 14 235304 161/839/0. OK} Nfd7 {0.81/10 10 130417} 11. h4 {0.75/16 58 935252 244/756/0. OK} b5 {0.87/10 12 199852} 12. g5 {0.89/12 12 194805 396/604/0} Bb7 {0.77/8 10 [#] 83633. Mistake (0.} (12... b4 $14) 13. Bd3 $1 $16 {1.52/13 25 388287 951/49/0. Strong} Nc5 {0.78/8 10 87216} 14. Bxh7+ {1.37/14 58 1028832 898/102/0. OK} Kxh7 {0.57/8 10 76748} 15. Qh5+ {1.49/13 58 1035576 942/58/0} Kg8 {0.55/5 10 7108} 16. O-O-O {1.44/12 13 239228 926/74/0. OK} Nc6 {0.88/7 10 136894. OK} 17. f5 {1.78/12 22 385683 987/13/0. OK} Nxd4 $4 {-0.47/8 10 [#] 69070. Loses game (Mate)} (17... Nxe5 $16) 18. g6 $1 {M+6/56 18 539995 1000/0/0. Strong} Ndb3+ {6.96/7 10 75946} 19. axb3 {M+5/63 19 552031 1000/0/0} Nxb3+ {10 M-5/7 52604} 20. cxb3 {M+4/79 19 668750 1000/0/0} fxg6 {10 M-3/3 4026} 21. fxg6 {M+3/83 21 745652 1000/0/0} Rf1 {10 M-2/2 540} 22. Qh7+ {M+2/113 27 165980 1000/0/0} Kf8 {10 M-1/3} 23. Qh8# {mate M+1/154 24 126719 1000/0/0.  Loses game:     ---       Black=1      Missed win:     ---       ---        Mistake:       ---       Black=1      Inaccurate:     ---       ---        OK:         White=5     Black=1      Best:         ---       ---        Strong:       White=2     ---        Brilliant:       ---       ---       .  Weighted Error Value: White=0.02/Black=0.50 Centipawn loss: w=2/b=32} 1-0

[/pgn]

[pgn][Event "banksia game"]
[Site "*"]
[Date "2022.12.31"]
[Round "1"]
[Board "28"]
[White "Dragon 3.1"]
[Black "Stockfish dev-20221221"]
[Result "0-1"]
[ECO "B56"]
[PlyCount "92"]
[Beauty "7220913778772"]
[EventDate "2022.??.??"]
[TimeControl "1"]

1. e4 c5 2. Nf3 Nc6 3. Nc3 d6 4. d4 cxd4 5. Nxd4 Nf6 6. f3 {B56: Sicilian} a6 {End of opening} 7. Be3 {0.82/8 15 172307. Stockfish dev-20221224-64656f85:} e5 {0.50/20 92 1461183 0/917/83} 8. Nb3 {0.64/9 10 87230} Be6 {0.45/17 15 299289 1/934/65} 9. Nd5 {0.69/10 10 171900} Bxd5 {0.41/16 13 273655 1/946/53. OK} 10. exd5 {0.56/10 10 104056} Ne7 {0.39/15 23 446985 1/954/45. OK} 11. c4 {0.65/11 10 78882. OK} g6 {0.24/14 22 424678 2/977/21. OK} 12. g4 {0.65/10 10 100171. Inaccurate} Bg7 {0.18/14 44 721220 2/983/15. OK} 13. Qd2 {0.48/9 12 184290. Inaccurate} O-O {-0.08/14 16 288294 9/988/3. Inaccurate} 14. Be2 {0.13/8 11 177750. Inaccurate} b5 {-0.53/15 17 366195 85/915/0. OK} 15. O-O {0.06/9 11 188104. OK} a5 {-0.33/21 177 3095166 31/968/1} 16. a4 {0.23/9 10 56780. Best} bxc4 {-0.32/14 10 184627 29/970/1. OK} 17. Bxc4 {0.03/10 10 127165} h5 {-0.30/16 11 213717 26/973/1} 18. h3 {0.00/9 10 155120. OK} Nh7 {-0.42/15 39 635655 46/954/0. OK} 19. Rad1 {0.33/9 12 203395. Inaccurate} f5 {-0.60/13 12 208284 115/885/0. Best} 20. Nc5 {0.07/8 10 87023. Inaccurate} f4 $1 {-0.84/14 12 240369 316/684/0. Strong} 21. Bf2 {-0.03/8 10 105192. OK} Ng5 {-0.88/12 14 214071 368/632/0. OK} 22. Bh4 {-0.18/9 10 109954. OK} Qb6 $1 {-0.99/14 15 256121 506/494/0. Strong} 23. Bxg5 {-0.40/8 10 78631. OK} Qxc5+ {-1.05/12 12 210660 594/406/0. OK} 24. Qf2 {-0.64/9 10 151443. Inaccurate} Qxc4 {-0.65/15 38 695375 140/860/0. OK} 25. Bxe7 {-0.77/10 10 94600. OK} hxg4 {-0.73/15 53 888046 194/806/0} 26. hxg4 {-0.45/10 13 213908} Rf7 {-0.79/12 10 133558 249/751/0. OK} 27. Bxd6 {-0.25/10 10 121385. OK} e4 {-0.84/14 13 241187 309/691/0. Best} 28. Qc5 {-0.25/11 10 133000} Qb3 {-1.14/15 49 759698 695/305/0} 29. Qc6 {-0.37/10 10 108425. Mistake (1.2)} (29. Qa3 $17) 29... Rd8 $1 $19 {-1.31/15 18 311898 848/152/0. Strong} 30. Bc5 {-0.54/11 10 199358. OK} e3 $1 {-1.36/15 22 374842 879/121/0. Strong} 31. Qe6 {-0.81/11 14 [#] 212747. Mistake (1.3)} (31. Qa6 $142) 31... Qc4 $1 $19 {-2.00/15 17 243899 996/4/0. Strong} 32. Bb6 $2 {-1.25/9 18 [#] 357025. Mistake (2.4)} (32. Rc1) 32... e2 $1 $19 {-4.11/12 15 202978 1000/0/0. Strong} 33. Rfe1 {-5.55/8 10 149791. OK} exd1=R {-4.36/12 46 817790 1000/0/0. OK} 34. Rxd1 {-5.75/8 10 89452} Qc2 {-4.44/13 48 833211 1000/0/0} 35. Re1 {-4.81/7 10 50703} Rb8 {-4.50/12 23 407942 1000/0/0. OK} 36. Rf1 {-5.42/7 10 61764. OK} Qb3 {-4.91/15 89 1466224 1000/0/0. OK} 37. Bf2 {-6.02/8 10 151187. Inaccurate} Qxf3 {-5.51/10 14 193692 1000/0/0. OK} 38. d6 {-6.48/7 10 51166. Inaccurate} Qh3 {-9.44/11 11 290255 1000/0/0} 39. Bb6 {-10.67/8 10 93438} Qg3+ {-40.26/14 11 296294 1000/0/0} 40. Kh1 {-11.79/5 10 8084} Rxb6 {M+11/23 10 201473 1000/0/0} 41. Rxf4 {-14.75/9 10 123756} Rxb2 {M+6/27 16 220810 1000/0/0} 42. Qxf7+ {10 M-5/8 45248} Kh7 {M+5/53 10 228672 1000/0/0. Black mates} 43. Rf2 {10 M-4/4 646} Rxf2 {M+4/55 17 201511 1000/0/0} 44. Qxf2 {10 M-3/1 352} Qxf2 {M+3/71 13 205892 1000/0/0} 45. g5 {10 M-2/3} Qf1+ {M+2/78 11 160535 1000/0/0} 46. Kh2 {10 M-1/2} Be5# {mate M+1/90 13 120683 1000/0/0.  Loses game:     ---       ---        Missed win:     ---       ---        Mistake:       White=3     ---        Inaccurate:     White=8     Black=1      OK:         White=11     Black=15     Best:         White=1     Black=2      Strong:       ---       Black=6      Brilliant:       ---       ---       .  Weighted Error Value: White=0.62/Black=0.02 Centipawn loss: w=40/b=3} 0-1

[/pgn]

[pgn][Event "banksia game"]
[Site "*"]
[Date "2022.12.31"]
[Round "1"]
[Board "42"]
[White "Dragon 3.1"]
[Black "Stockfish dev-20221221"]
[Result "0-1"]
[ECO "C50"]
[PlyCount "68"]
[Beauty "8041810374226"]
[EventDate "2022.??.??"]
[TimeControl "1"]

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. d3 Bc5 5. c3 {C50: Giuoco Pianissimo} d6 6. h3 O-O {End of opening} 7. O-O {0.37/1 23 6179} h6 {0.19/17 317 2308296 2/980/18. Stockfish dev-20221224-64656f85:} 8. b4 {0.63/9 13 181450} Bb6 {0.18/16 10 159788 3/981/16} 9. a4 {0.58/10 10 71291} a5 {0.18/15 12 223606 2/982/16} 10. b5 {0.55/9 10 90765. OK} Ne7 {0.15/18 27 556078 3/983/14. OK} 11. Ba3 {0.47/11 12 259717. Mistake (0.9)} (11. Nbd2 $11) 11... Ng6 $15 {-0.86/13 25 403906 360/640/0. OK} 12. Nbd2 {-0.17/9 11 128323. Inaccurate} Nf4 {-1.10/10 15 230905 674/326/0. OK} 13. d4 {0.06/8 10 [#] 61669} Re8 $1 {-1.00/12 31 500795 547/453/0. Strong} 14. Re1 {-0.02/8 10 [#] 129900. Inaccurate} Bxh3 $1 {-1.21/12 16 240731 785/215/0. Strong} 15. gxh3 {-0.36/9 10 123165} Nxh3+ {-1.35/11 17 254995 885/115/0} 16. Kg2 {-0.72/8 10 75120. Inaccurate} Nf4+ {-1.58/11 17 272004 963/37/0. OK} 17. Kh2 {-1.50/9 20 300809. Mistake (1.1)} (17. Kf1 $142) 17... Qd7 $19 {-1.74/12 15 262234 984/16/0. OK} 18. Bf1 {-1.94/9 10 149560. OK} d5 {-2.04/14 83 1370381 997/3/0. OK} 19. Nxe5 $2 {0.00/10 10 71742. Mistake (2.4)} (19. Bc5) 19... Rxe5 $19 {-2.52/12 13 215488 1000/0/0} 20. dxe5 {0.00/11 10 71919} Ng4+ {-3.88/13 20 305310 1000/0/0. OK} 21. Kg3 {0.00/11 10 47260. OK} Nh5+ {-4.49/11 26 387534 1000/0/0. Best} 22. Kg2 {-0.16/11 10 45250. Best} Qe6 {-4.48/12 21 328258 1000/0/0. OK} 23. exd5 {-1.94/9 10 150789} Qg6 {-6.52/11 12 181788 1000/0/0. OK} 24. Qb1 {-1.53/8 14 85012. Inaccurate} Qg5 {-7.67/12 28 511465 1000/0/0. OK} 25. Ne4 {-2.58/9 10 113922. Inaccurate} Ne3+ {M+14/13 30 718622 1000/0/0} 26. Kh2 {-8.26/8 10 75675} Qh4+ {M+11/23 19 255200 1000/0/0} 27. Bh3 {-10.67/9 10 60851. Black mates} Ng4+ {M+8/31 14 305506 1000/0/0} 28. Kg1 {-11.44/8 10 79892} Bxf2+ {M+7/33 10 344205 1000/0/0} 29. Kf1 {-10.94/9 12 146485} Nf4 {M+6/43 12 431830 1000/0/0} 30. Bxg4 {-9.72/8 10 77307} Qh2 {M+5/48 11 351783 1000/0/0} 31. Nf6+ {-11.79/9 10 76092} gxf6 {M+4/72 18 443375 1000/0/0} 32. Qg6+ {-11.88/10 10 62791} fxg6 {M+3/66 13 165431 1000/0/0} 33. Be6+ {-15.39/9 10 149715} Kh8 {M+2/106 14 132684 1000/0/0} 34. exf6 {10 M-1/1 2235} Qg1# {mate M+1/79 24 138983 1000/0/0.  Loses game:     ---       ---        Missed win:     ---       ---        Mistake:       White=3     ---        Inaccurate:     White=5     ---        OK:         White=3     Black=10     Best:         White=1     Black=1      Strong:       ---       Black=2      Brilliant:       ---       ---       .  Weighted Error Value: White=0.85/Black=0.00 Centipawn loss: w=81/b=1} 0-1

[/pgn]

Amazing.

Perfectionism in bullet games, then losing the advantage on longer time control.

Admin wrote:

Amazing.

Perfectionism in bullet games, then losing the advantage on longer time control.

The paradox is only caused by the terminology.

No chess engine plays perfect chess. And as a Type B Shannon chess engine. We know Stockfish can never play perfect chess.
And Stockfish could play a better move at micro bullet, and makes errors at deeper search depths.

For the simple fact that Stockfish is just guessing on all moves. A best guess, but Stockfish does not know what perfect play looks like.
Until it sees a solution with the EGTB, and EGTB probes.

mwyoung wrote:

In this test.

I played 50 games of chess at 1 second per game, or just microsecond of thinking time per move.

The games were played between Stockfish 15 DEV vs Dragon 3.1. They are the strongest chess engines on the planet.

banksia game  2022

                                 
1   Stockfish dev-20221221  +346  +38/=12/-0 88.00%   44.0/50
2   Dragon 3.1              -346  +0/=12/-38 12.00%    6.0/50


It is not shocking that Stockfish won this micro bullet match.

What is shocking is the level of play Stockfish demonstrated at this micro bullet match.

Here are the centipawn analysis results from this match.

Dragon 3.1:   38/4/22/14/2/34/62/70/133/28/43/24/40/33/46/5/46/25/79/41/85/66/30/1/27/76/6/38/29/37/32/22/37/29/38/25/19/4/6/29/50/38/44/29/22/61/5/31/2/0  => Average=0.35

Dragon played at a grandmaster level of play for a grandmaster playing a chess game at 3 minutes a move, or at standard FIDE time controls.
When playing chess at 1 second per game. Dragon 3.1 had a average move error rate of .35 Centipawns per move.

As a frame of reference I took the best games of GM Magnus Carlsen. And the games were all wins, and played in the last 7 years.
So this is not GM Carlsen's average error rate, but the best error rate of his best cherry picked games.

GM Magnus Carlsen's best error rate is .12 Centipawns per move. In his best 50 games. Played in the last 7 years. And that were all wins at 3 minutes per move. Or FIDE standard time controls.

Carlsen Magnus:   2/10/10/21/14/23/23/4/30/22/49/8/4/15/18/35/6/8/16/5/11/4/9/13/14/7/5/12/7/20/12/27/7/14/1/13/8/2/18/7/11/10/10/6/4/10/2/4/13/7  => Average=0.12

Stockfish at 1 second per game scored an incredible .07 centipawn loss per move average in 50 games.

Stockfish dev-20221221:   7/4/10/14/1/4/2/2/41/6/3/12/1/17/2/4/3/4/17/5/0/11/1/1/3/26/6/2/2/10/14/2/4/28/9/3/2/3/5/8/2/4/6/5/9/8/4/7/2/0  => Average=0.07

Stockfish demonstrated a level of play much better then GM Magnus Carlsen can demonstrate in his best games. On his best days, from the last 7 years. Played at 3 minutes a move, or FIDE standard time controls.

At 1 Second Per Game.

I think that it may be easier to have smaller error rate when you play against weaker opponents.

I wonder what happens if you give Dragon to play against some weaker engine at 1 second per game and if it is not going to have a smaller centipawn loss per move average relative to the centipawn loss per move against Stockfish.

Uri Blass wrote:

mwyoung wrote:

In this test.

I played 50 games of chess at 1 second per game, or just microsecond of thinking time per move.

The games were played between Stockfish 15 DEV vs Dragon 3.1. They are the strongest chess engines on the planet.

banksia game  2022

                                 
1   Stockfish dev-20221221  +346  +38/=12/-0 88.00%   44.0/50
2   Dragon 3.1              -346  +0/=12/-38 12.00%    6.0/50


It is not shocking that Stockfish won this micro bullet match.

What is shocking is the level of play Stockfish demonstrated at this micro bullet match.

Here are the centipawn analysis results from this match.

Dragon 3.1:   38/4/22/14/2/34/62/70/133/28/43/24/40/33/46/5/46/25/79/41/85/66/30/1/27/76/6/38/29/37/32/22/37/29/38/25/19/4/6/29/50/38/44/29/22/61/5/31/2/0  => Average=0.35

Dragon played at a grandmaster level of play for a grandmaster playing a chess game at 3 minutes a move, or at standard FIDE time controls.
When playing chess at 1 second per game. Dragon 3.1 had a average move error rate of .35 Centipawns per move.

As a frame of reference I took the best games of GM Magnus Carlsen. And the games were all wins, and played in the last 7 years.
So this is not GM Carlsen's average error rate, but the best error rate of his best cherry picked games.

GM Magnus Carlsen's best error rate is .12 Centipawns per move. In his best 50 games. Played in the last 7 years. And that were all wins at 3 minutes per move. Or FIDE standard time controls.

Carlsen Magnus:   2/10/10/21/14/23/23/4/30/22/49/8/4/15/18/35/6/8/16/5/11/4/9/13/14/7/5/12/7/20/12/27/7/14/1/13/8/2/18/7/11/10/10/6/4/10/2/4/13/7  => Average=0.12

Stockfish at 1 second per game scored an incredible .07 centipawn loss per move average in 50 games.

Stockfish dev-20221221:   7/4/10/14/1/4/2/2/41/6/3/12/1/17/2/4/3/4/17/5/0/11/1/1/3/26/6/2/2/10/14/2/4/28/9/3/2/3/5/8/2/4/6/5/9/8/4/7/2/0  => Average=0.07

Stockfish demonstrated a level of play much better then GM Magnus Carlsen can demonstrate in his best games. On his best days, from the last 7 years. Played at 3 minutes a move, or FIDE standard time controls.

At 1 Second Per Game.

I think that it may be easier to have smaller error rate when you play against weaker opponents.

I wonder what happens if you give Dragon to play against some weaker engine at 1 second per game and if it is not going to have a smaller centipawn loss per move average relative to the centipawn loss per move against Stockfish.

Feel free to post those results.

I played the top 2 chess engines on the planet in this match.

mwyoung wrote:

In this test.

I played 50 games of chess at 1 second per game, or just microsecond of thinking time per move.

The games were played between Stockfish 15 DEV vs Dragon 3.1. They are the strongest chess engines on the planet.

banksia game  2022

                                 
1   Stockfish dev-20221221  +346  +38/=12/-0 88.00%   44.0/50
2   Dragon 3.1              -346  +0/=12/-38 12.00%    6.0/50


It is not shocking that Stockfish won this micro bullet match.

What is shocking is the level of play Stockfish demonstrated at this micro bullet match.





At 1 Second Per Game.

I agree with you about his strengh, but instead it scales badly on powerful hardware! I'm testing my engine blitz 5 0 on Playchess.com using only a Core i5 4 cores (2 MN/s)
against ThreadRipper and Xeon up to 128 threads and I'm first! It seems that Stockfish isn't much stronger with powerful hardware..

MY tests with ProteusSF: http://t.ly/iToU-

Alex-Microsmeta wrote:

mwyoung wrote:

In this test.

I played 50 games of chess at 1 second per game, or just microsecond of thinking time per move.

The games were played between Stockfish 15 DEV vs Dragon 3.1. They are the strongest chess engines on the planet.

banksia game  2022

                                 
1   Stockfish dev-20221221  +346  +38/=12/-0 88.00%   44.0/50
2   Dragon 3.1              -346  +0/=12/-38 12.00%    6.0/50


It is not shocking that Stockfish won this micro bullet match.

What is shocking is the level of play Stockfish demonstrated at this micro bullet match.





At 1 Second Per Game.

I agree with you about his strengh, but instead it scales badly on powerful hardware! I'm testing my engine blitz 5 0 on Playchess.com using only a Core i5 4 cores (2 MN/s)
against ThreadRipper and Xeon up to 128 threads and I'm first! It seems that Stockfish isn't much stronger with powerful hardware..

MY tests with ProteusSF: http://t.ly/iToU-

I have done the same testing. And I have found it is better to have fewer faster cores. Then many cores running Stockfish. My Laptop running Stockfish is faster then my Threadripper.

That is why I try and show the data on Stockfish scaling. But people only see and like NPS with Stockfish. Even to the point of paying huge money for Cloud Servers running Stockfish. To get that warm fuzzy feeling.