Развернуть для просмотра

Владимир001 · 08.12.2020, 01:03

PGN - http://kirill-kryukov.com/chess/disc...e.php?id=47503

Код:

Code:
CCRL 40/15 Rating List - Custom engine selection
1195608 games played by 2746 programs, run by 23 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.
Computed on December 5, 2020 with Bayeselo based on 1'195'608 games
Tested by CCRL team, 2005-2020, http://ccrl.chessdom.com/ccrl/4040/

Rank                 Engine                   Elo   +    -   Score  AvOp  Games
1 tomitankChess 4.0 64-bit                2818  +22  -22  49.6%   +0.2   667
  tomitankChess 4.2 64-bit                2816  +26  -26  50.4%   -6.8   449
  tomitankChess 3.0 64-bit                2786  +16  -16  49.2%   +1.1  1330
  tomitankChess 2.0 64-bit                2657  +18  -18  51.9%  -15.3  1105
  tomitankChess 2.1 64-bit                2637  +22  -22  47.7%  +13.2   710
  tomitankChess 1.5 64-bit                2590  +23  -23  50.2%   -1.6   611
  tomitankChess 1.4 64-bit                2556  +23  -23  52.7%  -18.9   632

Владимир001 · 08.12.2020, 01:05

PGN - http://kirill-kryukov.com/chess/disc...e.php?id=47502

Код:

Code:
CCRL 40/15 Rating List - Custom engine selection
1195608 games played by 2746 programs, run by 23 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.
Computed on December 5, 2020 with Bayeselo based on 1'195'608 games
Tested by CCRL team, 2005-2020, http://ccrl.chessdom.com/ccrl/4040/

Rank                 Engine                   Elo   +    -   Score  AvOp  Games
1 Ethereal 12.75 64-bit 4CPU              3389  +24  -24  55.0%  -26.6   474
  Ethereal 12.25 64-bit 4CPU              3366  +22  -22  49.3%   +5.3   544
  Ethereal 12.00 64-bit 4CPU              3352  +22  -22  50.9%   -3.3   564
  Ethereal 11.75 64-bit 4CPU              3347  +21  -21  47.3%  +15.0   614
  Ethereal 11.53 64-bit 4CPU              3325  +23  -22  52.3%  -13.1   542
  Ethereal 11.25 64-bit 4CPU              3299  +21  -22  46.3%  +24.1   588
  Ethereal 11.00 64-bit 4CPU              3298  +19  -19  47.0%  +18.1   778
  Ethereal 10.55 64-bit 4CPU              3274  +24  -25  49.8%   -2.2   488
  Ethereal 10.00 64-bit 4CPU              3188  +20  -20  53.4%  -22.9   788
  Ethereal 9.00 64-bit 4CPU               3050  +23  -23  46.8%  +18.4   570
  Ethereal 8.62 64-bit 4CPU               2930  +22  -22  44.2%  +35.8   600

Владимир001 · 08.12.2020, 01:07

PGN - http://kirill-kryukov.com/chess/disc...e.php?id=47501

Код:

CCRL 40/15 Rating List - Custom engine selection
1195608 games played by 2746 programs, run by 23 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.
Computed on December 5, 2020 with Bayeselo based on 1'195'608 games
Tested by CCRL team, 2005-2020, http://ccrl.chessdom.com/ccrl/4040/

Rank                 Engine                   Elo   +    -   Score  AvOp  Games
1 ShashChess 15.0 64-bit                  3479  +19  -19  75.3% -168.7   951
  Stockfish 12 64-bit                     3474  +17  -16  75.3% -167.3  1249
  SugaR XPro NN 1.0 64-bit                3419  +15  -15  67.4% -111.4  1443
  ShashChess 11.0 64-bit                  3404  +16  -16  67.2% -107.8  1212

Владимир001 · 08.12.2020, 01:09

PGN - http://kirill-kryukov.com/chess/disc...e.php?id=47500

Код:

CCRL 40/15 Rating List - Custom engine selection
1195608 games played by 2746 programs, run by 23 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.
Computed on December 5, 2020 with Bayeselo based on 1'195'608 games
Tested by CCRL team, 2005-2020, http://ccrl.chessdom.com/ccrl/4040/

Rank                 Engine                   Elo   +    -   Score  AvOp  Games
1 Mr Bob 0.9.0 64-bit                     2842  +29  -28  58.7%  -64.3   418
  Mr Bob 0.8.0 64-bit                     2651  +27  -27  56.1%  -42.2   442
  Mr Bob 0.7.1 64-bit                     2558  +28  -28  51.1%   -8.9   408

Владимир001 · 08.12.2020, 11:08

SMOOTH MOVES

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
GBSelect2019.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 36 rounds
All engines 64-bit where available

Final Standings

21.5 - RookieMonster 1.8.1 64-bit
21.0 - KnightX 2.4 64-bit
21.0 - Betsabe II 2020
19.5 - Velvet 1.0.0 64-bit
19.5 - Raven 1.10 64-bit
19.0 - MadChess 2.2 64-bit
18.0 - Supernova 2.2.1 64-bit
15.5 - paulchen332 0.1.1 64-bit
13.5 - Ghost 3.1 64-bit
11.5 - Drofa 2.0.0 64-bit

The complete tournament pgn (zipped) can be downloaded here:
http://kirill-kryukov.com/chess/disc...e.php?id=47505

ChessMan · 08.12.2020, 14:13

Ryzen 3900X @4.2GHz
GUI: CuteChess 1.2
book: 8moves_v3.pgn
1 core, concurrency 10, hash 16 MB
Ponder=off
TC: 10+0.1

Use NNUE = Hybrid

Код:

   # PLAYER                                     :  RATING  ERROR  PLAYED   (%)    W     D    L  D(%)  CFS(%)
   1 CiChess_x64_021220_E_AVX2_sparse           :    3205      3    2371  51.5  205  2033  133  85.7     100
   2 CiChess_x64_021220_E_SSE2_popcnt_sparse    :    3195      3    2371  48.5  133  2033  205  85.7     ---

White advantage = 20.73 +/- 2.37
Draw rate (equal opponents) = 87.13 % +/- 0.70

PGN: https://pixeldrain.com/u/cN2ePkQh

Владимир001 · 08.12.2020, 17:05

Season 2 Halfway (games 90/180)

Leela JH 94-80 leads

Код:

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)
+1 Lc0-j94-80          :     141     54    17.5      30  58.3      60    6   23    1  76.7
 2 Stockfish 301120    :     131     53    17.0      30  56.7      77    5   24    1  80.0
 3 Dragon              :     101     50    15.5      30  51.7      60    5   21    4  70.0
 4 Lc0 T60-66390       :      91     51    15.0      30  50.0      60    3   24    3  80.0
 5 Leelenstein 15      :      81     51    14.5      30  48.3      97    2   25    3  83.3
 6 SF Classical        :       0     59    10.5      30  35.0     ---    3   15   12  50.0

White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 79.29 % +/- 5.55

MEDIANS

Код:

Name                    Games    Depth       NPS           Nodes
-------------------------------------------------------------------------------------
Lc0-j94-80          :      30     16           55473      5065857
Lc0 T60-66390       :      30     19           70660      7208976
Stockfish 301120    :      30     51        33070855   1293358107
Leelenstein 15      :      30     18          103723      8984805
Dragon              :      30     33        30127975   1350177482
SF Classical        :      30     43        41284840   1656495903

Владимир001 · 08.12.2020, 17:18

Crystal 051220 wins JCER Test new engines, 2020.12.05 - 2020.12.06
Temp 3' + 3" GUI-Fritz 17 Ggames

Владимир001 · 09.12.2020, 13:14

Match: Match of CorChess NNUE 4CPU against Lc0 66513

Развернуть для просмотра

Lc0-version: v0.26.3
Lc0-options : -Backend=cudnn -NNCacheSize=2000000 -MaxOutOfOrderEvalsFactor=2.4 -MaxCollisionEvents=256 -TwoFoldDraws=true
MLH-settings S19 SuFi: -MovesLeftMaxEffect=0.2 -MovesLeftSlope=0.004 -MovesLeftThreshold=0.0 -MovesQuadraticFactor=0.0 -MovesScaledFactor=1.0
CorChess-version: Corchess NNUE 1.3 081120 x64 bmi2 (08.11.2020)
CorChess-options: threads=4 hash=4096 MB, NNUE=true, net=nn-c3ca321c51c9.nnue
Time control: 5min + 5s
Hardware: notebook with mobile i7-7700HQ 4x2.8GHz (with HT), mobile Geforce GTX 1050 TI 4GB
Book: Balsa_Special.pgn, 10 plies, sequentially, colors reversed
Tablebases: Syzygy 5-men on SSD
Speed: around 20k nodes per move for J92/J94 // T60: 1.5k nps, J92/J94: 1.0k nps, CorChess NNUE 4CPU: 4112k nps
Adjudication: Draw at 50 moves rule, 3-fold repetition, after 200 moves, at 50 moves with eval = 0.00 ; 5-men Gaviota tablebase adjudication
Software: Arena 3.5.1, ponder=off
Elo rating: Ordo 1.2.6
Comments: 66513 is one of the first T60 nets with a bigger net size of 383x30 (same size as J94/J92). ERET test results included in the table.

Код:

   # PLAYER                        :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)  ERET test
   1 Corchess NNUE 1.3_081120_4CPU :     0.0   ----   522.5     900  58.1     100  150  745    5  82.8  97/111
   2 Lc0.net.JH.J92-280            :   -46.2   24.2    43.5     100  43.5      58    1   85   14  85.0  80/111
   3 Lc0.net.JH.J94-80             :   -49.8   25.6    43.0     100  43.0      58    1   84   15  84.0  80/111
   4 Lc0.net.66350                 :   -53.4   25.4    42.5     100  42.5      58    0   85   15  85.0  82/111
   5 Lc0.net.66167                 :   -57.0   27.3    42.0     100  42.0      50    1   82   17  82.0  82/111
   6 Lc0.net.JH.J94-40             :   -57.0   26.5    42.0     100  42.0      57    0   84   16  84.0  79/111
   7 Lc0.net.JH.J94-60             :   -60.6   27.0    41.5     100  41.5      50    1   81   18  81.0  80/111
   8 Lc0.net.66513                 :   -60.6   27.5    41.5     100  41.5      57    0   83   17  83.0  82/111    <== NEW
   9 Lc0.net.65981                 :   -64.2   27.9    41.0     100  41.0      57    1   80   19  80.0  82/111
  10 Lc0.net.JH.J92-330            :   -67.9   28.7    40.5     100  40.5     ---    0   81   19  81.0  79/111

https://pixeldrain.com/u/PXSwUuGP

[свернуть]

Владимир001 · 09.12.2020, 13:18

SF-dev vs. LC0 24x320 nets

Развернуть для просмотра

Hardware: i5-8400 CPU @ 2.80GHz, 6 cores, 32GB, gpu RTX 2060, 6GB
Network IDs: T60B.7-105, T60B.7-135, 62948, 64255, 65162, 65297, 65373, 65374, 65726, 65762, 65830, 65889, 65939, 65988, 66053, 66103, 66148, 66234, 66275, 66314, 66353, 66381, 66430, 66471, 66511
LC0-version: v0.26.0
SF-version: SF-dev 20092708 (NNUE hybrid)
Software: Cutechess-CLI
Time control: LC0: 1,500 nodes/move, SF-dev: 300,000 nodes/move @ 4 Threads, 512MB Hash
Book: Chad's 8-ply book with 10000 positions, order=sequential
Tablebases: 6-men
Adjudication: -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
Ordo: -a 0 -A "Stockfish.dev" -W -D -U "0,1,2,3,4,5,6,7,8,9,10" -s 2000 -n 1
Comment:

Код:

   # PLAYER                :  RATING  ERROR    POINTS  PLAYED   (%)  CFS(%)      W       D       L  D(%)
+  1 lc0.net.66511         :    41.3    3.2   11142.0   20000    56      90   5655   10974    3371    55
   2 lc0.net.65939         :    38.2    3.3   11058.0   20000    55      59   5592   10932    3476    55
   3 lc0.net.66353         :    37.7    3.2   11043.5   20000    55      54   5608   10871    3521    54
   4 lc0.net.66471         :    37.5    3.2   11037.5   20000    55      58   5505   11065    3430    55
   5 lc0.net.66430         :    37.0    3.2   11024.5   20000    55      58   5501   11047    3452    55
   6 lc0.net.66314         :    36.5    3.2   11011.5   20000    55      60   5518   10987    3495    55
   7 lc0.net.66148         :    35.9    3.2   10994.5   20000    55      52   5498   10993    3509    55
   8 lc0.net.66381         :    35.8    3.3   10991.0   20000    55      56   5488   11006    3506    55
   9 lc0.net.65889         :    35.5    3.2   10982.0   20000    55      61   5529   10906    3565    55
  10 lc0.net.65988         :    34.8    3.3   10963.0   20000    55      50   5474   10978    3548    55
  11 lc0.net.66234         :    34.8    3.3   10962.5   20000    55      51   5495   10935    3570    55
  12 lc0.net.66103         :    34.7    3.2   10961.0   20000    55      55   5449   11024    3527    55
  13 lc0.net.65762         :    34.4    3.3   10952.0   20000    55      57   5494   10916    3590    55
  14 lc0.net.66275         :    34.0    3.2   10940.5   20000    55      58   5406   11069    3525    55
  15 lc0.net.65830         :    33.5    3.2   10927.5   20000    55      68   5457   10941    3602    55
  16 lc0.net.66053         :    32.4    3.2   10897.5   20000    54      71   5363   11069    3568    55
  17 lc0.net.65726         :    31.1    3.3   10862.0   20000    54      66   5317   11090    3593    55
  18 lc0.net.65297         :    30.2    3.1   10836.0   20000    54      59   5374   10924    3702    55
  19 lc0.net.65373         :    29.7    3.3   10822.0   20000    54      85   5360   10924    3716    55
  20 lc0.net.65162         :    27.2    3.2   10754.0   20000    54      93   5255   10998    3747    55
  21 lc0.net.65374         :    23.7    3.3   10657.5   20000    53      90   5127   11061    3812    55
  22 lc0.net.T60B.7-135    :    20.8    3.2   10576.5   20000    53      63   4994   11165    3841    56
  23 lc0.net.T60B.7-105    :    20.0    3.2   10556.0   20000    53     100   4984   11144    3872    56
  24 lc0.net.62948         :     5.9    3.2   10163.0   20000    51      83   4400   11526    4074    58
  25 lc0.net.64255         :     3.6    3.2   10100.0   20000    51      99   4463   11274    4263    56
- 26 Stockfish.dev         :     0.0   ----  228784.5  500000    46     ---  90875  275819  133306    55

White advantage = 57.67 +/- 0.32
Draw rate (equal opponents) = 57.43 % +/- 0.07

PGN:
https://pixeldrain.com/u/Xgn9sAip

[свернуть]

Владимир001 · 09.12.2020, 15:35

asmFishWCP_20019-09-30 Vs Stockfish 051220 NNUE Off

Time control : 3 minutes
Computer : X 3450, 12 Gb, Windows 10
Book : No book for both engines
Engine setting : Default setting for both engines (Stockfish 051220 minus nn-62ef826d1a6d.nnue)
Hashtable : 1024 mb
Threads : 4
Ponder : Off
GUI : ChessProgram 17
EGTB : 6 men
Games : 100
-----------------------------------------------------------------------------------------------
1. asmFishWCP_20019-09-30_popcnt +20/=65/-15 52.50% 52.5/100 17

2. Stockfish 051220 (NNUE: Off) +15/=65/-20 47.50% 47.5/100 -17

Detail information>>>

Владимир001 · 10.12.2020, 05:21

Match: T60 66550 vs Stockfish 20111513 in fixed nodes

Развернуть для просмотра

LC0 version: 0.26.3
LC0 options: cudnn-fp16, 1 thread, default for the rest
SF options: 1 thread, 1GB Hash.
Time control: Leela: 1 kn/move, SF dev: 300Kn/move
Hardware: RTX 2070, i7-7700
Book: Balsa 2724 openings
Tablebases: 6-man TB for both.
Adjudication: 6-man TB, -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
Software:cutechess-cli
Comments: T60 nets now have the same size as J9x nets. 66550 is in the same strength level as J9x nets.

Код:

Leela:  1 thread, 1Kn/move
SF dev: 1 thread, 300Kn/move

Balsa 2724
   # PLAYER                 :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)     W      D     L
   1 lc0.net.J92-330.1k     :       3      6   2747.0    5448  50.4      51   985   3524   939
   2 lc0.net.J94-80.1k      :       3      6   2746.5    5448  50.4      50   988   3517   943
   3 lc0.net.J96-44.1k      :       3      6   2746.5    5448  50.4      55   978   3537   933
   4 lc0.net.J94-60.1k      :       2      6   2742.5    5448  50.3      80   981   3523   944
-  5 SF_dev20111513.300k    :       0   ----  16288.0   32688  49.8      74  5722  21132  5834
+  6 lc0.net.66550.1k       :      -2      5   2711.0    5448  49.8      56   947   3528   973
   7 lc0.net.J96-28.1k      :      -2      6   2706.5    5448  49.7     ---   955   3503   990

https://pixeldrain.com/u/a67nRjtw

[свернуть]

Владимир001 · 10.12.2020, 05:25

Match: Lc0 26.3 vs Stockfish 12 avx2 (4T)
Time control: 5s + 0.5s

Развернуть для просмотра

Hardware: RTX 2060, Ryzen-9 4900H
Opening book: noob_3moves.pgn
Stockfish: threads=4, hash=64, MoveOverhead=0
Lc0: cuda-fp16 threads=2, minibatch-size=60, mlh, move-overhead=0
mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0
TBs and adj.: syzygy 5-men, resign movecount=5 score=600, draw movenumber=30 movecount=5 score=5, time forfeit disabled
Comments: 66510 added. MLH seems to slightly hurt 66510 performance in this set up (-8 elo, CFS of 93%), and lowers the draw rate.
Games: https://gofile.io/d/9B0R6F

Код:

# PLAYER                      : RATING  ERROR  PLAYED   (%)   CFS    W     D    L    D(%)
1 Stockfish 12 4CPU           :    0.0   ----   10000  55.36  100  1885  7302  813  73.02
2 lc0-263-cuda-j96-28         :  -30.1    8.3    2000  45.73   58   178  1473  349  73.65
3 lc0-263-cuda-j92-330        :  -31.3    7.9    2000  45.55   52   183  1456  361  72.80
4 lc0-263-cuda-j96-44         :  -31.6    7.9    2000  45.50   97   163  1494  343  74.70
5 lc0-263-cuda-66510-mlh-off  :  -43.9    8.3    2000  43.77   93   142  1467  391  73.35
6 lc0-263-cuda-66510          :  -51.9    8.7    2000  42.65  ---   147  1412  441  70.60

White advantage = 8.59 +/- 1.95
Draw rate (equal opponents) = 74.85 % +/- 0.43

[свернуть]

Владимир001 · 10.12.2020, 05:31

Цитата: Сообщение от Владимир001

Season 2 Halfway (games 90/180)

Leela JH 94-80 leads

Код:

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)
+1 Lc0-j94-80          :     141     54    17.5      30  58.3      60    6   23    1  76.7
 2 Stockfish 301120    :     131     53    17.0      30  56.7      77    5   24    1  80.0
 3 Dragon              :     101     50    15.5      30  51.7      60    5   21    4  70.0
 4 Lc0 T60-66390       :      91     51    15.0      30  50.0      60    3   24    3  80.0
 5 Leelenstein 15      :      81     51    14.5      30  48.3      97    2   25    3  83.3
 6 SF Classical        :       0     59    10.5      30  35.0     ---    3   15   12  50.0

White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 79.29 % +/- 5.55

MEDIANS

Код:

Name                    Games    Depth       NPS           Nodes
-------------------------------------------------------------------------------------
Lc0-j94-80          :      30     16           55473      5065857
Lc0 T60-66390       :      30     19           70660      7208976
Stockfish 301120    :      30     51        33070855   1293358107
Leelenstein 15      :      30     18          103723      8984805
Dragon              :      30     33        30127975   1350177482
SF Classical        :      30     43        41284840   1656495903

Season 2 Games 120/180 (40 games each)

Развернуть для просмотра

Book: TCEC 19
Time control 1800 + 30
Hardware: i9-10980xe, 3080+2070, 64GB

Код:

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)
+1 Lc0-j94-80          :      44     44    23.0      40  57.5      74    7   32    1  80.0
 2 Stockfish 301120    :      22     42    21.5      40  53.8      67    5   33    2  82.5
 3 Dragon              :       7     42    20.5      40  51.3      50    5   31    4  77.5
 4 Lc0 T60-66390       :       7     42    20.5      40  51.3      81    4   33    3  82.5
 5 Leelenstein 15      :     -22     42    18.5      40  46.3      87    2   33    5  82.5
 6 SF Classical        :     -59     45    16.0      40  40.0     ---    5   22   13  55.0

White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 80.40 % +/- 4.28

MEDIANS

Код:

Name                    Games    Depth       NPS           Nodes
-------------------------------------------------------------------------------------
Lc0-j94-80          :      40     16           56536      5152264
Lc0 T60-66390       :      40     19           69227      7391820
Stockfish 301120    :      40     52        33281425   1356389439
Leelenstein 15      :      40     19           98858      9631280
Dragon              :      40     34        30526844   1421988630
SF Classical        :      40     44        41891118   1725879927

https://www.twitch.tv/navratil25

[свернуть]

Владимир001 · 11.12.2020, 09:23

83RD AMATEUR SERIES (Division 1)

Intel i7 4770k Quad
ChessGUI
256mb hash each where possible
3-4-5 piece tablebases
Ponder off
ChessGUI.cgb book
40 moves in 16 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 1CPU (64-bit where possible)

Final Standings

34.0 - Stockfish 12 64-bit
32.0 - ShashChess 15.0 64-bit
29.5 - SugaR XPro NN 1.0 64-bit
24.0 - Nemorino 6.00 64-bit
23.5 - Ethereal 12.75 64-bit
20.5 - SlowChess Blitz Classic 2.4 64-bit
19.5 - Igel 2.8.0 64-bit
18.0 - Booot 6.4 64-bit
17.0 - Xiphos 0.6 64-bit
17.0 - RubiChess 1.9 64-bit
16.5 - RofChade 2.3 64-bit
12.5 - Schooner 2.2 64-bit

The complete tournament pgn (zipped) can be downloaded here:
http://kirill-kryukov.com/chess/disc...e.php?id=47512

Владимир001 · 11.12.2020, 10:22

Spark 1.0 wins JCER League group E1 edition 41, 2020.12.06 - 2020.12.09

Temp 3' + 3" 306 games GUI-Fritz 15
Promotion to the League 6: Spark 1.0, Weiss 1.1.
Relegation to Group E2: Invictus r305, GnuCheese 1.00, Rhetoric 1.4.3.

Владимир001 · 11.12.2020, 19:40

Match: PhoenixStein 1.1 vs SF-12 4CPU

Развернуть для просмотра

Time control: 2m + 2s
Hardware: Ryzen-7 3750H, GTX 1660 Ti
Openings: TCEC SuFi 17-18-19 (150 pos.)
Lc0 26.3 cudnn-fp16: nncache=2000000, threads=2, minibatch-size=64, move-overhead=0
Stockfish 12 modern: threads=4, hash=256, MoveOverhead=0
Benchmarks startpos (3500ms = average time/move): SF-12 4CPU = 2450k, PhoenixS. = 5.6k, LS-15 = 6.8k
TBs and adj.: syzygy 5-men, draw 5 moves 5cp move 30, resign (two-sided) 5 moves 600cp, (time forfeit disabled)
Comments: New net by @jjosh about 50 elo behind LS-15 on tcec data, even if reaching higher depth/move.
Games: https://gofile.io/d/jKGwba

Код:

# PLAYER                   : RATING  ERROR  PLAYED  (%)   CFS   W    D   L    D(%)
1 stockfish-12-4cpu        :    0.0   ----    600  57.67   99  186  320  94  53.33
2 lc0-263-ls-15            :  -35.5   29.0    300  45.50   99   58  157  85  52.33
3 lc0-263-PhoenixStein-1.1 :  -86.4   26.6    300  39.17  ---   36  163 101  54.33

White advantage = 120.23 +/- 10.19
Draw rate (equal opponents) = 63.71 % +/- 2.29

Engine                 Depth   MIDG   EARLY    ENDG   LATE
lc0-PhoenixStein-1.1   12.95  13.94 | 14.50 | 11.73 | 10.16
lc0-ls-15              11.63  12.70 | 12.85 |  9.98 |  9.28
stockfish-12-4cpu      29.40  26.03 | 26.48 | 32.41 | 43.03

[свернуть]

Владимир001 · 12.12.2020, 09:59

JCER CUP 2020, Final - match no 5 Raubfisch - Stockfish

Promotion: Raubfisch X44, Temp 60' + 10" GUI - Fritz 17

Владимир001 · 12.12.2020, 10:14

Match: T73 vs SF-12
Time control: 2s + 0.2s

Развернуть для просмотра

Hardware: Ryzen-7 3750H, GTX 1660 Ti
Openings: TCEC SuFi 10-19 (497 pos.) + Balsa_v500 (500 pos.)
Lc0 26.3 cudnn-fp16: threads=1, minibatch-size=64, mlh tcec-19=on, move-overhead=0
mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0
Stockfish 12: threads=1, hash=16, MoveOverhead=0
TBs and adj.: syzygy 5-men, draw 5 moves 5cp move 30, resign (two-sided) 5 moves 600cp, (time forfeit disabled)
Comments: 730730 added. It looks like we have a trend here! Last three nets tested were: 528 -> 620 -> 730. Maybe T73 will end up catching 703810? Also, the draw rate went suddently up.
Games: https://gofile.io/d/G1Fzhk

Код:

# PLAYER            : RATING  ERROR  PLAYED   (%)   CFS    W     D     L   D(%)
1 lc0-263-703810    :    1.8    9.4    1994  50.25   65   469  1066   459  53.46
2 stockfish-12-1cpu :    0.0   ----   13958  51.75   89  3547  7352  3059  52.67
3 lc0-263-730730    :   -6.1   10.0    1994  49.15   83   435  1090   469  54.66
4 lc0-263-730620    :  -13.2   10.9    1994  48.17   69   462   997   535  50.00
5 lc0-263-730517    :  -16.8   10.8    1994  47.67   53   412  1077   505  54.01
6 lc0-263-730370    :  -17.3   10.7    1994  47.59   53   438  1022   534  51.25
7 lc0-263-730528    :  -17.9    9.8    1994  47.52   54   425  1045   524  52.41
8 lc0-263-730315    :  -18.6   11.1    1994  47.42  ---   418  1055   521  52.91

White advantage = 55.44 +/- 1.92
Draw rate (equal opponents) = 54.21 % +/- 0.42

[свернуть]

Владимир001 · 12.12.2020, 10:17

Match: PR963 (betamcts) vs baseline

Развернуть для просмотра

LC0 version (both): deeabd7
LC0 options (both): threads=1, minibatch-size=32, cpuct=1.3, cpuct-at-root=2.5, april-factor=0, april-factor-parent=0, cpuct-factor=2.815, cpuct-factor-at-root=2.815, fpu-value=0.3, fpu-strategy-at-root=absolute, policy-softmax-temp=1.4, minimum-kldgain-per-node=0.00002, smart-pruning-factor=0, moves-left-max-effect=0.1, moves-left-threshold=0, moves-left-slope=0.007, moves-left-quadratic-factor=1, move-overhead-ms=0
LC0 options (betamcts): betamcts-level=2, betamcts-trust=3.12, betamcts-prior=32.6, lcb-percentile=0.306
LC0 options (baseline): betamcts-level=0, betamcts-trust=0, betamcts-prior=0, lcb-percentile=0
Time control: visits=10000 (kld)
Net: 703810
Book: custom-3k (1-3 ply)
Hardware: gtx970 + i5-4570
Tablebases: syzygy 5-man
Adjudication: -draw movenumber=50 movecount=5 score=10 -resign movecount=5 score=700 twosided=true
Software: cutechess-cli
Comment: training simulation with betamcts using kld (no noise or temp). based on most recent betamcts tune

Код:

   # PLAYER                 :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W     D    L  D(%)
   1 lc0-703810-betamcts    :    28.1    9.1  1080.0    2000    54     100  431  1298  271    65
   2 lc0-703810             :     0.0   ----   920.0    2000    46     ---  271  1298  431    65

White advantage = 4.59 +/- 4.72
Draw rate (equal opponents) = 65.52 % +/- 1.08

https://pixeldrain.com/u/LJdfoe5h

[свернуть]

Владимир001 · 12.12.2020, 14:43

Season 2 Games 150/180 (50 games each)

Book: TCEC 19
Time control 1800 + 30
Hardware: i9-10980xe, 3080+2070, 64GB

Развернуть для просмотра

Код:

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)    W    D    L  D(%)
+1 Lc0-j94-80          :      54     38    29.5      50  59.0      79   10   39    1  78.0
 2 Stockfish 301120    :      30     37    27.5      50  55.0      79    7   41    2  82.0
 3 Lc0 T60-66390       :       6     37    25.5      50  51.0      73    5   41    4  82.0
 4 Dragon              :     -12     37    24.0      50  48.0      58    5   38    7  76.0
 5 Leelenstein 15      :     -18     36    23.5      50  47.0      92    2   43    5  86.0
 6 SF Classical        :     -60     40    20.0      50  40.0     ---    5   30   15  60.0

White advantage = 0.00 +/- 0.00
Draw rate (equal opponents) = 82.02 % +/- 3.82

Код:

MEDIANS
Name                    Games    Depth       NPS           Nodes
-------------------------------------------------------------------------------------
Lc0-j94-80          :      50     16           58893      5271577
Lc0 T60-66390       :      50     19           68530      7730220
Stockfish 301120    :      50     51        33506389   1405890838
Leelenstein 15      :      50     19           99301     10044227
Dragon              :      50     34        31203871   1489645830
SF Classical        :      50     45        42387410   1783251642

https://www.twitch.tv/navratil25

[свернуть]

Владимир001 · 12.12.2020, 14:50

66512 evaluation is finished. The next net is 66592. First comparison in 3 days.

Hardware: i5-8400 CPU @ 2.80GHz, 6 cores, 32GB, gpu RTX 2060, 6GB
Gauntlet: SF-dev vs. LC0 384x30b nets

Развернуть для просмотра

Network IDs: 66512
LC0-version: v0.26.3
LC0-options: minibatch-size=90
SF-version: SF-dev 20120516
SF-options: 4 Threads, 512MB Hash
Software: Cutechess-CLI
Time control: LC0: 1,500 nodes/move, SF-dev: 500,000 nodes/move
Book: Chad's 8-ply book with 10000 positions, order=sequential
Tablebases: 6-men
Adjudication: -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
Ordo: -a 0 -A "Stockfish.dev" -W -D -U "0,1,2,3,4,5,6,7,8,9,10" -s 2000 -n 1

Код:

# PLAYER                :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)     W      D     L  D(%)
1 lc0.net.66512         :     8.8    2.9  10247.0   20000    51     100  3935  12624  3441    63
2 Stockfish.dev         :     0.0   ----   9753.0   20000    49     ---  3441  12624  3935    63

White advantage = 51.35 +/- 1.52
Draw rate (equal opponents) = 65.12 % +/- 0.35

https://docs.google.com/spreadsheets...it?usp=sharing

[свернуть]

Владимир001 · 13.12.2020, 11:31

Match: Lc0 26.3 vs Stockfish 12 avx2 (4T)
Time control: 5s + 0.5s
Hardware: RTX 2060, Ryzen-9 4900H
Opening book: noob_3moves.pgn

Развернуть для просмотра

Stockfish: threads=4, hash=64, MoveOverhead=0
Lc0: cuda-fp16, threads=2, minibatch-size=60, mlh, move-overhead=0
mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0
TBs and adj.: syzygy 5-men, resign movecount=5 score=600, draw movenumber=30 movecount=5 score=5, time forfeit disabled
Comments: 66555 added and joins the leading group. 66555, J92-330, and the J96's are within 1.5 elo of each other. I again tested mlh on/off, and this time mlh=on got +5.5 elo. Maybe "mlh tcec-19" suits better 30x384 nets (66555) than 24x320 nets (66510).
Games: https://gofile.io/d/6w2eaM

Код:

# PLAYER                 : RATING  ERROR  PLAYED   (%)   CFS    W      D     L   D(%)
1 Stockfish 12 4CPU      :    0.0   ----   14000  55.20  100  2613  10231  1156  73.08
2 lc0-263-j96-28         :  -30.1    8.5    2000  45.73   58   178   1473   349  73.65
3 lc0-263-66555          :  -31.1    7.0    2000  45.58   51   180   1463   357  73.15
4 lc0-263-j92-330        :  -31.3    7.5    2000  45.55   53   183   1456   361  72.80
5 lc0-263-j96-44         :  -31.6    7.5    2000  45.50   83   163   1494   343  74.70
6 lc0-263-66555-mlh-off  :  -36.6    7.3    2000  44.80   89   163   1466   371  73.30
7 lc0-263-66510-mlh-off  :  -43.9    8.4    2000  43.77   91   142   1467   391  73.35
8 lc0-263-66510          :  -51.9    7.9    2000  42.65  ---   147   1412   441  70.60

White advantage = 6.91 +/- 1.40
Draw rate (equal opponents) = 74.76 % +/- 0.36

[свернуть]

Владимир001 · 14.12.2020, 03:14

JCER Cup 2020 - Match Zeus 10.2 - Fisherov 0.93c

Promotion: Zeus 10.2, Temp 60' + 10" GUI - Fritz 17

Владимир001 · 14.12.2020, 03:15

Cfish 081220 wins JCER Test new engines (8-9.12.2020)

Temp 3' + 3" 112 games GUI-Fritz 17

Владимир001 · 14.12.2020, 03:53

Match: lc0 nets vs Stockfish - 100 games TC:40/20 rep

Развернуть для просмотра

LC0 version: (Cudnn variant) v0.26.3; 1 Thread
LC0 options: Default (i.e. Kiudi bonus)
mlh_settings-[MovesLeftMaxEffect=0.2, MovesLeftThreshold=0.0, MovesLeftSlope=0.004, MovesLeftConstantFactor=0.0, MovesLeftScaledFactor=1.0,
MovesLeftQuadraticFactor=0.0]
SF options: SF12(nn-82215), Threads 4, Hash 4GB
Hardware: CPU i7-4790 with 16 GByte RAM + GTX1650
Time control: 40/20 rep (i.e. 40 moves in 20 minutes repeating)
Book: Custom 5-move book, played from both sides (Balanced_Openings_50.pgn)
Tablebases: 5 piece syzygy for both engines
Adjudication: 5-man Gaviota TB adjudication, draw movenumber=50 movecount=3 score=0.00, resign movecount=5 score=900
Software: Arena 3.5.1
Comment: Tracking progress of leela networks at relatively longer time control

Код:

     
# PLAYER                   :  RATING  ERROR  POINTS  PLAYED    W    D    L   (%)  CFS(%)  D(%)
1 Leela v0.26.3 J92-330    :    21.2   22.9    53.0     100    9   88    3    53      97    88
2 Stockfish_12_x64_bmi2    :     0.0   ----    97.5     200    9  177   14    49      62    89
3 Leela v0.26.3 66512      :    -3.5   22.7    49.5     100    5   89    6    50     ---    89 (new)

White advantage = 26.39 +/- 8.29
Draw rate (equal opponents) = 91.70 % +/- 2.52

[свернуть]

Владимир001 · 14.12.2020, 03:58

V3.0.0 (219) тест

Vafra 3.0.0 против Houdini 6; игра в шахматы с медленным контролем времени на ноутбуке HP Pavilion dv7, использующем Intel® Core ™ i7, с четырьмя гиперпоточными ядрами, оба движка думали только о своем собственном ходу, т.е. без размышлений; оба движка использовали шесть потоков, оба движка использовали одну и ту же книгу открытия, Fritz 16, минимум в 3 партиях до 8 ходов; У Вафры был доступ ко всем таблицам эндшпиля на 5 фигурки , у Гудини не было доступа к базам финальных игр; Vafra 3.0.0 - это модифицированная Vafra 2.17.0 с NNUE, Vafra использовала nn-62ef826d1a6d.nnue NNUE.

http://www.jurjevic.org.uk/chess/gam...udini_022.html

Владимир001 · 14.12.2020, 20:04

3 min 2 core frıtz15 gui syzgyz 3-4-5 nnue i7 3820 4ghz play 112 games

Код:

                                  1          2          3          4          5          6          7          8          
1   Cfish 131220 64 POPCNT NUMA   ********** ½½½½       ½½½½       ½½½½       ½½½½       ½½½½       ½½1½       1111        16.5/28
2   Eman 6.61 64-bit SSE2         ½½½½       ********** ½½½½       ½½½½       ½½½½       ½½½½       ½½½½       1111        16.0/28
3   CF EXT 081220 x64 SSE2 N      ½½½½       ½½½½       ********** ½½½½       ½½½½       ½½½½       ½½½½       111½        15.5/28
4  Cfish 131220 64 POPCNT NUMAc  ½½½½       ½½½½       ½½½½       ********** ½½1½       ½½½½       ½½½½       1½½½        15.0/28  204.00
5  CF-EXT-021220   64 POPCNT        ½½½½       ½½½½       ½½½½       ½½0½       ********** ½½½1       ½½½½       1½1½        15.0/28  198.75
6   CiChess 081220 x64 SSE2 N     ½½½½       ½½½½       ½½½½       ½½½½       ½½½0       ********** ½½½½       ½1½1        14.5/28  192.50
7   Raubfisch X44a_nn_sl-modern   ½½0½       ½½½½       ½½½½       ½½½½       ½½½½       ½½½½       ********** ½½11        14.5/28  191.75
8   Houdidit 6.03 Pro x64-popc    0000       0000       000½       0½½½       0½0½       ½0½0       ½½00       **********   5.0/28

https://pixeldrain.com/u/82KWAxqf

Владимир001 · 15.12.2020, 02:23

Dragon 1.0 wins 2 League JCER edition 41, 2020.12.10 - 2020.12.13

Temp 3' + 3" 264 games GUI-Fritz 17

Владимир001 · 15.12.2020, 10:05

Hardware: i5-8400 CPU @ 2.80GHz, 6 cores, 32GB, gpu RTX 2060, 6GB
Gauntlet: SF-dev vs. LC0 384x30b nets

Развернуть для просмотра

Network IDs: 66512, 66592
LC0-version: v0.26.3
LC0-options: minibatch-size=90
SF-version: SF-dev 20120516
SF-options: 4 Threads, 512MB Hash
Software: Cutechess-CLI
Time control: LC0: 1,500 nodes/move, SF-dev: 500,000 nodes/move
Book: Chad's 8-ply book with 10000 positions, order=sequential
Tablebases: 6-men
Adjudication: -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000
Ordo: -a 0 -A "Stockfish.dev" -W -D -U "0,1,2,3,4,5,6,7,8,9,10" -s 2000 -n 1
Comment:

Код:

   # PLAYER           :  RATING  ERROR   POINTS  PLAYED   (%)  CFS(%)     W      D     L  D(%)
+  1 lc0.net.66592    :    14.1    3.0  10394.5   20000    52      99  4200  12389  3411    62
+  2 lc0.net.66512    :     8.8    2.9  10247.0   20000    51     100  3935  12624  3441    63
-  3 Stockfish.dev    :     0.0   ----  19358.5   40000    48     ---  6852  25013  8135    63

White advantage = 51.81 +/- 1.06
Draw rate (equal opponents) = 64.56 % +/- 0.25

PGN:
2 files:
https://pixeldrain.com/l/XpNKBcPH#item=0

[свернуть]

08.12.2020, 11:08	#665
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков SMOOTH MOVES Intel i5 Quad ChessGUI 256mb hash each 3-4-5 piece tablebases Ponder off GBSelect2019.cgb book 40 moves in 25 minutes repeating (adapted for the CCRL) 4 cycles 36 rounds All engines 64-bit where available Final Standings 21.5 - RookieMonster 1.8.1 64-bit 21.0 - KnightX 2.4 64-bit 21.0 - Betsabe II 2020 19.5 - Velvet 1.0.0 64-bit 19.5 - Raven 1.10 64-bit 19.0 - MadChess 2.2 64-bit 18.0 - Supernova 2.2.1 64-bit 15.5 - paulchen332 0.1.1 64-bit 13.5 - Ghost 3.1 64-bit 11.5 - Drofa 2.0.0 64-bit The complete tournament pgn (zipped) can be downloaded here: http://kirill-kryukov.com/chess/disc...e.php?id=47505

08.12.2020, 14:13	#666
ChessMan Super Moderator Регистрация: 04.01.2017 Сообщений: 3,629 Сказал(а) спасибо: 2,879 Поблагодарили 6,463 раз(а) в 2,331 сообщениях Репутация: 354	Re: Разные матчи и турниры движков Ryzen 3900X @4.2GHz GUI: CuteChess 1.2 book: 8moves_v3.pgn 1 core, concurrency 10, hash 16 MB Ponder=off TC: 10+0.1 Use NNUE = Hybrid Код: # PLAYER : RATING ERROR PLAYED (%) W D L D(%) CFS(%) 1 CiChess_x64_021220_E_AVX2_sparse : 3205 3 2371 51.5 205 2033 133 85.7 100 2 CiChess_x64_021220_E_SSE2_popcnt_sparse : 3195 3 2371 48.5 133 2033 205 85.7 --- White advantage = 20.73 +/- 2.37 Draw rate (equal opponents) = 87.13 % +/- 0.70 PGN: https://pixeldrain.com/u/cN2ePkQh

08.12.2020, 17:18	#668
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Crystal 051220 wins JCER Test new engines, 2020.12.05 - 2020.12.06 Temp 3' + 3" GUI-Fritz 17 Ggames

09.12.2020, 13:14	#669
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: Match of CorChess NNUE 4CPU against Lc0 66513 Развернуть для просмотра Lc0-version: v0.26.3 Lc0-options : -Backend=cudnn -NNCacheSize=2000000 -MaxOutOfOrderEvalsFactor=2.4 -MaxCollisionEvents=256 -TwoFoldDraws=true MLH-settings S19 SuFi: -MovesLeftMaxEffect=0.2 -MovesLeftSlope=0.004 -MovesLeftThreshold=0.0 -MovesQuadraticFactor=0.0 -MovesScaledFactor=1.0 CorChess-version: Corchess NNUE 1.3 081120 x64 bmi2 (08.11.2020) CorChess-options: threads=4 hash=4096 MB, NNUE=true, net=nn-c3ca321c51c9.nnue Time control: 5min + 5s Hardware: notebook with mobile i7-7700HQ 4x2.8GHz (with HT), mobile Geforce GTX 1050 TI 4GB Book: Balsa_Special.pgn, 10 plies, sequentially, colors reversed Tablebases: Syzygy 5-men on SSD Speed: around 20k nodes per move for J92/J94 // T60: 1.5k nps, J92/J94: 1.0k nps, CorChess NNUE 4CPU: 4112k nps Adjudication: Draw at 50 moves rule, 3-fold repetition, after 200 moves, at 50 moves with eval = 0.00 ; 5-men Gaviota tablebase adjudication Software: Arena 3.5.1, ponder=off Elo rating: Ordo 1.2.6 Comments: 66513 is one of the first T60 nets with a bigger net size of 383x30 (same size as J94/J92). ERET test results included in the table. Код: # PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) ERET test 1 Corchess NNUE 1.3_081120_4CPU : 0.0 ---- 522.5 900 58.1 100 150 745 5 82.8 97/111 2 Lc0.net.JH.J92-280 : -46.2 24.2 43.5 100 43.5 58 1 85 14 85.0 80/111 3 Lc0.net.JH.J94-80 : -49.8 25.6 43.0 100 43.0 58 1 84 15 84.0 80/111 4 Lc0.net.66350 : -53.4 25.4 42.5 100 42.5 58 0 85 15 85.0 82/111 5 Lc0.net.66167 : -57.0 27.3 42.0 100 42.0 50 1 82 17 82.0 82/111 6 Lc0.net.JH.J94-40 : -57.0 26.5 42.0 100 42.0 57 0 84 16 84.0 79/111 7 Lc0.net.JH.J94-60 : -60.6 27.0 41.5 100 41.5 50 1 81 18 81.0 80/111 8 Lc0.net.66513 : -60.6 27.5 41.5 100 41.5 57 0 83 17 83.0 82/111 <== NEW 9 Lc0.net.65981 : -64.2 27.9 41.0 100 41.0 57 1 80 19 80.0 82/111 10 Lc0.net.JH.J92-330 : -67.9 28.7 40.5 100 40.5 --- 0 81 19 81.0 79/111 https://pixeldrain.com/u/PXSwUuGP [свернуть]

09.12.2020, 15:35	#671
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков asmFishWCP_20019-09-30 Vs Stockfish 051220 NNUE Off Time control : 3 minutes Computer : X 3450, 12 Gb, Windows 10 Book : No book for both engines Engine setting : Default setting for both engines (Stockfish 051220 minus nn-62ef826d1a6d.nnue) Hashtable : 1024 mb Threads : 4 Ponder : Off GUI : ChessProgram 17 EGTB : 6 men Games : 100 ----------------------------------------------------------------------------------------------- 1. asmFishWCP_20019-09-30_popcnt +20/=65/-15 52.50% 52.5/100 17 2. Stockfish 051220 (NNUE: Off) +15/=65/-20 47.50% 47.5/100 -17 Detail information>>>

10.12.2020, 05:21	#672
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: T60 66550 vs Stockfish 20111513 in fixed nodes Развернуть для просмотра LC0 version: 0.26.3 LC0 options: cudnn-fp16, 1 thread, default for the rest SF options: 1 thread, 1GB Hash. Time control: Leela: 1 kn/move, SF dev: 300Kn/move Hardware: RTX 2070, i7-7700 Book: Balsa 2724 openings Tablebases: 6-man TB for both. Adjudication: 6-man TB, -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000 Software:cutechess-cli Comments: T60 nets now have the same size as J9x nets. 66550 is in the same strength level as J9x nets. Код: Leela: 1 thread, 1Kn/move SF dev: 1 thread, 300Kn/move Balsa 2724 # PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L 1 lc0.net.J92-330.1k : 3 6 2747.0 5448 50.4 51 985 3524 939 2 lc0.net.J94-80.1k : 3 6 2746.5 5448 50.4 50 988 3517 943 3 lc0.net.J96-44.1k : 3 6 2746.5 5448 50.4 55 978 3537 933 4 lc0.net.J94-60.1k : 2 6 2742.5 5448 50.3 80 981 3523 944 - 5 SF_dev20111513.300k : 0 ---- 16288.0 32688 49.8 74 5722 21132 5834 + 6 lc0.net.66550.1k : -2 5 2711.0 5448 49.8 56 947 3528 973 7 lc0.net.J96-28.1k : -2 6 2706.5 5448 49.7 --- 955 3503 990 https://pixeldrain.com/u/a67nRjtw [свернуть]

10.12.2020, 05:25	#673
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: Lc0 26.3 vs Stockfish 12 avx2 (4T) Time control: 5s + 0.5s Развернуть для просмотра Hardware: RTX 2060, Ryzen-9 4900H Opening book: noob_3moves.pgn Stockfish: threads=4, hash=64, MoveOverhead=0 Lc0: cuda-fp16 threads=2, minibatch-size=60, mlh, move-overhead=0 mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0 TBs and adj.: syzygy 5-men, resign movecount=5 score=600, draw movenumber=30 movecount=5 score=5, time forfeit disabled Comments: 66510 added. MLH seems to slightly hurt 66510 performance in this set up (-8 elo, CFS of 93%), and lowers the draw rate. Games: https://gofile.io/d/9B0R6F Код: # PLAYER : RATING ERROR PLAYED (%) CFS W D L D(%) 1 Stockfish 12 4CPU : 0.0 ---- 10000 55.36 100 1885 7302 813 73.02 2 lc0-263-cuda-j96-28 : -30.1 8.3 2000 45.73 58 178 1473 349 73.65 3 lc0-263-cuda-j92-330 : -31.3 7.9 2000 45.55 52 183 1456 361 72.80 4 lc0-263-cuda-j96-44 : -31.6 7.9 2000 45.50 97 163 1494 343 74.70 5 lc0-263-cuda-66510-mlh-off : -43.9 8.3 2000 43.77 93 142 1467 391 73.35 6 lc0-263-cuda-66510 : -51.9 8.7 2000 42.65 --- 147 1412 441 70.60 White advantage = 8.59 +/- 1.95 Draw rate (equal opponents) = 74.85 % +/- 0.43 [свернуть]

11.12.2020, 09:23	#675
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков 83RD AMATEUR SERIES (Division 1) Intel i7 4770k Quad ChessGUI 256mb hash each where possible 3-4-5 piece tablebases Ponder off ChessGUI.cgb book 40 moves in 16 minutes repeating (adapted for the CCRL) 4 cycles 44 rounds All engines 1CPU (64-bit where possible) Final Standings 34.0 - Stockfish 12 64-bit 32.0 - ShashChess 15.0 64-bit 29.5 - SugaR XPro NN 1.0 64-bit 24.0 - Nemorino 6.00 64-bit 23.5 - Ethereal 12.75 64-bit 20.5 - SlowChess Blitz Classic 2.4 64-bit 19.5 - Igel 2.8.0 64-bit 18.0 - Booot 6.4 64-bit 17.0 - Xiphos 0.6 64-bit 17.0 - RubiChess 1.9 64-bit 16.5 - RofChade 2.3 64-bit 12.5 - Schooner 2.2 64-bit The complete tournament pgn (zipped) can be downloaded here: http://kirill-kryukov.com/chess/disc...e.php?id=47512

11.12.2020, 10:22	#676
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Spark 1.0 wins JCER League group E1 edition 41, 2020.12.06 - 2020.12.09 Temp 3' + 3" 306 games GUI-Fritz 15 Promotion to the League 6: Spark 1.0, Weiss 1.1. Relegation to Group E2: Invictus r305, GnuCheese 1.00, Rhetoric 1.4.3.

11.12.2020, 19:40	#677
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: PhoenixStein 1.1 vs SF-12 4CPU Развернуть для просмотра Time control: 2m + 2s Hardware: Ryzen-7 3750H, GTX 1660 Ti Openings: TCEC SuFi 17-18-19 (150 pos.) Lc0 26.3 cudnn-fp16: nncache=2000000, threads=2, minibatch-size=64, move-overhead=0 Stockfish 12 modern: threads=4, hash=256, MoveOverhead=0 Benchmarks startpos (3500ms = average time/move): SF-12 4CPU = 2450k, PhoenixS. = 5.6k, LS-15 = 6.8k TBs and adj.: syzygy 5-men, draw 5 moves 5cp move 30, resign (two-sided) 5 moves 600cp, (time forfeit disabled) Comments: New net by @jjosh about 50 elo behind LS-15 on tcec data, even if reaching higher depth/move. Games: https://gofile.io/d/jKGwba Код: # PLAYER : RATING ERROR PLAYED (%) CFS W D L D(%) 1 stockfish-12-4cpu : 0.0 ---- 600 57.67 99 186 320 94 53.33 2 lc0-263-ls-15 : -35.5 29.0 300 45.50 99 58 157 85 52.33 3 lc0-263-PhoenixStein-1.1 : -86.4 26.6 300 39.17 --- 36 163 101 54.33 White advantage = 120.23 +/- 10.19 Draw rate (equal opponents) = 63.71 % +/- 2.29 Engine Depth MIDG EARLY ENDG LATE lc0-PhoenixStein-1.1 12.95 13.94 \| 14.50 \| 11.73 \| 10.16 lc0-ls-15 11.63 12.70 \| 12.85 \| 9.98 \| 9.28 stockfish-12-4cpu 29.40 26.03 \| 26.48 \| 32.41 \| 43.03 [свернуть]

12.12.2020, 09:59	#678
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков JCER CUP 2020, Final - match no 5 Raubfisch - Stockfish Promotion: Raubfisch X44, Temp 60' + 10" GUI - Fritz 17

12.12.2020, 10:14	#679
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: T73 vs SF-12 Time control: 2s + 0.2s Развернуть для просмотра Hardware: Ryzen-7 3750H, GTX 1660 Ti Openings: TCEC SuFi 10-19 (497 pos.) + Balsa_v500 (500 pos.) Lc0 26.3 cudnn-fp16: threads=1, minibatch-size=64, mlh tcec-19=on, move-overhead=0 mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0 Stockfish 12: threads=1, hash=16, MoveOverhead=0 TBs and adj.: syzygy 5-men, draw 5 moves 5cp move 30, resign (two-sided) 5 moves 600cp, (time forfeit disabled) Comments: 730730 added. It looks like we have a trend here! Last three nets tested were: 528 -> 620 -> 730. Maybe T73 will end up catching 703810? Also, the draw rate went suddently up. Games: https://gofile.io/d/G1Fzhk Код: # PLAYER : RATING ERROR PLAYED (%) CFS W D L D(%) 1 lc0-263-703810 : 1.8 9.4 1994 50.25 65 469 1066 459 53.46 2 stockfish-12-1cpu : 0.0 ---- 13958 51.75 89 3547 7352 3059 52.67 3 lc0-263-730730 : -6.1 10.0 1994 49.15 83 435 1090 469 54.66 4 lc0-263-730620 : -13.2 10.9 1994 48.17 69 462 997 535 50.00 5 lc0-263-730517 : -16.8 10.8 1994 47.67 53 412 1077 505 54.01 6 lc0-263-730370 : -17.3 10.7 1994 47.59 53 438 1022 534 51.25 7 lc0-263-730528 : -17.9 9.8 1994 47.52 54 425 1045 524 52.41 8 lc0-263-730315 : -18.6 11.1 1994 47.42 --- 418 1055 521 52.91 White advantage = 55.44 +/- 1.92 Draw rate (equal opponents) = 54.21 % +/- 0.42 [свернуть]

12.12.2020, 10:17	#680
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: PR963 (betamcts) vs baseline Развернуть для просмотра LC0 version (both): deeabd7 LC0 options (both): threads=1, minibatch-size=32, cpuct=1.3, cpuct-at-root=2.5, april-factor=0, april-factor-parent=0, cpuct-factor=2.815, cpuct-factor-at-root=2.815, fpu-value=0.3, fpu-strategy-at-root=absolute, policy-softmax-temp=1.4, minimum-kldgain-per-node=0.00002, smart-pruning-factor=0, moves-left-max-effect=0.1, moves-left-threshold=0, moves-left-slope=0.007, moves-left-quadratic-factor=1, move-overhead-ms=0 LC0 options (betamcts): betamcts-level=2, betamcts-trust=3.12, betamcts-prior=32.6, lcb-percentile=0.306 LC0 options (baseline): betamcts-level=0, betamcts-trust=0, betamcts-prior=0, lcb-percentile=0 Time control: visits=10000 (kld) Net: 703810 Book: custom-3k (1-3 ply) Hardware: gtx970 + i5-4570 Tablebases: syzygy 5-man Adjudication: -draw movenumber=50 movecount=5 score=10 -resign movecount=5 score=700 twosided=true Software: cutechess-cli Comment: training simulation with betamcts using kld (no noise or temp). based on most recent betamcts tune Код: # PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) 1 lc0-703810-betamcts : 28.1 9.1 1080.0 2000 54 100 431 1298 271 65 2 lc0-703810 : 0.0 ---- 920.0 2000 46 --- 271 1298 431 65 White advantage = 4.59 +/- 4.72 Draw rate (equal opponents) = 65.52 % +/- 1.08 https://pixeldrain.com/u/LJdfoe5h [свернуть]

12.12.2020, 14:50	#682
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков 66512 evaluation is finished. The next net is 66592. First comparison in 3 days. Hardware: i5-8400 CPU @ 2.80GHz, 6 cores, 32GB, gpu RTX 2060, 6GB Gauntlet: SF-dev vs. LC0 384x30b nets Развернуть для просмотра Network IDs: 66512 LC0-version: v0.26.3 LC0-options: minibatch-size=90 SF-version: SF-dev 20120516 SF-options: 4 Threads, 512MB Hash Software: Cutechess-CLI Time control: LC0: 1,500 nodes/move, SF-dev: 500,000 nodes/move Book: Chad's 8-ply book with 10000 positions, order=sequential Tablebases: 6-men Adjudication: -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000 Ordo: -a 0 -A "Stockfish.dev" -W -D -U "0,1,2,3,4,5,6,7,8,9,10" -s 2000 -n 1 Код: # PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) 1 lc0.net.66512 : 8.8 2.9 10247.0 20000 51 100 3935 12624 3441 63 2 Stockfish.dev : 0.0 ---- 9753.0 20000 49 --- 3441 12624 3935 63 White advantage = 51.35 +/- 1.52 Draw rate (equal opponents) = 65.12 % +/- 0.35 https://docs.google.com/spreadsheets...it?usp=sharing [свернуть]

13.12.2020, 11:31	#683
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Match: Lc0 26.3 vs Stockfish 12 avx2 (4T) Time control: 5s + 0.5s Hardware: RTX 2060, Ryzen-9 4900H Opening book: noob_3moves.pgn Развернуть для просмотра Stockfish: threads=4, hash=64, MoveOverhead=0 Lc0: cuda-fp16, threads=2, minibatch-size=60, mlh, move-overhead=0 mlh tcec-19: moves-left-max-effect=0.2, moves-left-threshold=0, moves-left-slope=0.004, moves-left-scaled-factor=1, moves-left-quadratic-factor=0, moves-left-constant-factor=0 TBs and adj.: syzygy 5-men, resign movecount=5 score=600, draw movenumber=30 movecount=5 score=5, time forfeit disabled Comments: 66555 added and joins the leading group. 66555, J92-330, and the J96's are within 1.5 elo of each other. I again tested mlh on/off, and this time mlh=on got +5.5 elo. Maybe "mlh tcec-19" suits better 30x384 nets (66555) than 24x320 nets (66510). Games: https://gofile.io/d/6w2eaM Код: # PLAYER : RATING ERROR PLAYED (%) CFS W D L D(%) 1 Stockfish 12 4CPU : 0.0 ---- 14000 55.20 100 2613 10231 1156 73.08 2 lc0-263-j96-28 : -30.1 8.5 2000 45.73 58 178 1473 349 73.65 3 lc0-263-66555 : -31.1 7.0 2000 45.58 51 180 1463 357 73.15 4 lc0-263-j92-330 : -31.3 7.5 2000 45.55 53 183 1456 361 72.80 5 lc0-263-j96-44 : -31.6 7.5 2000 45.50 83 163 1494 343 74.70 6 lc0-263-66555-mlh-off : -36.6 7.3 2000 44.80 89 163 1466 371 73.30 7 lc0-263-66510-mlh-off : -43.9 8.4 2000 43.77 91 142 1467 391 73.35 8 lc0-263-66510 : -51.9 7.9 2000 42.65 --- 147 1412 441 70.60 White advantage = 6.91 +/- 1.40 Draw rate (equal opponents) = 74.76 % +/- 0.36 [свернуть]

14.12.2020, 03:14	#684
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков JCER Cup 2020 - Match Zeus 10.2 - Fisherov 0.93c Promotion: Zeus 10.2, Temp 60' + 10" GUI - Fritz 17

14.12.2020, 03:15	#685
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Cfish 081220 wins JCER Test new engines (8-9.12.2020) Temp 3' + 3" 112 games GUI-Fritz 17

14.12.2020, 03:58	#687
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков V3.0.0 (219) тест Vafra 3.0.0 против Houdini 6; игра в шахматы с медленным контролем времени на ноутбуке HP Pavilion dv7, использующем Intel® Core ™ i7, с четырьмя гиперпоточными ядрами, оба движка думали только о своем собственном ходу, т.е. без размышлений; оба движка использовали шесть потоков, оба движка использовали одну и ту же книгу открытия, Fritz 16, минимум в 3 партиях до 8 ходов; У Вафры был доступ ко всем таблицам эндшпиля на 5 фигурки , у Гудини не было доступа к базам финальных игр; Vafra 3.0.0 - это модифицированная Vafra 2.17.0 с NNUE, Vafra использовала nn-62ef826d1a6d.nnue NNUE. http://www.jurjevic.org.uk/chess/gam...udini_022.html

15.12.2020, 02:23	#689
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Dragon 1.0 wins 2 League JCER edition 41, 2020.12.10 - 2020.12.13 Temp 3' + 3" 264 games GUI-Fritz 17

15.12.2020, 10:05	#690
Владимир001 Эксперт Регистрация: 19.12.2019 Сообщений: 7,484 Сказал(а) спасибо: 4,655 Поблагодарили 12,178 раз(а) в 5,475 сообщениях Репутация: 259	Re: Разные матчи и турниры движков Hardware: i5-8400 CPU @ 2.80GHz, 6 cores, 32GB, gpu RTX 2060, 6GB Gauntlet: SF-dev vs. LC0 384x30b nets Развернуть для просмотра Network IDs: 66512, 66592 LC0-version: v0.26.3 LC0-options: minibatch-size=90 SF-version: SF-dev 20120516 SF-options: 4 Threads, 512MB Hash Software: Cutechess-CLI Time control: LC0: 1,500 nodes/move, SF-dev: 500,000 nodes/move Book: Chad's 8-ply book with 10000 positions, order=sequential Tablebases: 6-men Adjudication: -draw movenumber=50 movecount=5 score=8 -resign movecount=5 score=1000 Ordo: -a 0 -A "Stockfish.dev" -W -D -U "0,1,2,3,4,5,6,7,8,9,10" -s 2000 -n 1 Comment: Код: # PLAYER : RATING ERROR POINTS PLAYED (%) CFS(%) W D L D(%) + 1 lc0.net.66592 : 14.1 3.0 10394.5 20000 52 99 4200 12389 3411 62 + 2 lc0.net.66512 : 8.8 2.9 10247.0 20000 51 100 3935 12624 3441 63 - 3 Stockfish.dev : 0.0 ---- 19358.5 40000 48 --- 6852 25013 8135 63 White advantage = 51.81 +/- 1.06 Draw rate (equal opponents) = 64.56 % +/- 0.25 PGN: 2 files: https://pixeldrain.com/l/XpNKBcPH#item=0 [свернуть]

Здесь присутствуют: 3 (пользователей: 0 , гостей: 3)