FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Discussion of computer chess matches and engine tournaments.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
fastgm
Posts: 552
Joined: Mon Aug 19, 2013 4:57 pm
Contact:

FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Post by fastgm » Tue Jun 30, 2020 8:46 pm

New: 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)
Moves Left Head (MLH) Parameters: MovesLeftMaxEffect=0.2, MovesLeftThreshold=0, MovesLeftSlope=0.007, MovesLeftQuadraticFactor=0, MovesLeftScaledFactor=1.0, MovesLeftConstantFactor=0

Code: Select all

   # Engine                            :    Elo  Error   Games    Points    (%)      W      D      L     D(%)   CFS
 -------------------------------------------------------------------------------------------------------------------
   1 Lc0 0.25.1 LS 15                  :   3481     11    2500    1971.5   78.86   1511    921     68   36.84    81
   2 Lc0 0.24.1 LS 14.3                :   3474     11    2500    1956.0   78.24   1476    960     64   38.40    77
   3 Lc0 0.24.1 LS 14.2                :   3469      9    3250    2472.5   76.08   1799   1347    104   41.45    89
   4 Lc0 0.24.0 LS 14                  :   3460      9    3500    2677.0   76.49   1964   1426    110   40.74    92
   5 Lc0k 0.23.2 256x20-T40-1541       :   3452      8    4250    3145.5   74.01   2226   1839    185   43.27    75
   6 Lc0 0.24.1 LS 13                  :   3447      9    3000    2187.5   72.92   1508   1359    133   45.30    86
   7 Stockfish 200502 C0               :   3441      8    2500    1301.5   52.06    383   1837    280   73.48    94
   8 Lc0 0.25.1 63681                  :   3430     10    2500    1852.0   74.08   1308   1088    104   43.52    52
   9 Lc0 0.24.1 384x30-T60-3010        :   3430      9    3500    2565.0   73.29   1782   1566    152   44.74    54
  10 Fat Fritz 1.1                     :   3429      8    4000    2921.0   73.03   2066   1710    224   42.75    76
  11 Stockfish 11 C0                   :   3426      5    5500    2790.5   50.74    728   4125    647   75.00    56
  12 Lc0 0.22.0 42850                  :   3425      7    4500    3173.5   70.52   2142   2063    295   45.84    51
  13 Lc0 0.22.0 T40B.4-160             :   3425      7    4500    3173.0   70.51   2131   2084    285   46.31    82
  14 Stockfish 11                      :   3421      6    4000    2157.5   53.94    777   2761    462   69.03    78
  15 Lc0 0.25.1 63305                  :   3416     10    2500    1823.5   72.94   1246   1155     99   46.20    57
  16 Lc0 0.25.1 63305 (Kayra4)         :   3415     10    2500    1820.0   72.80   1239   1162     99   46.48    90
  17 Lc0k 0.23.2 62172                 :   3405     10    2500    1827.0   73.08   1256   1142    102   45.68    53
  18 Lc0 LS-LS 12.2                    :   3405      8    3750    2680.5   71.48   1809   1743    198   46.48    53
  19 Lc0 0.25.1 703042                 :   3404     10    2500    1784.5   71.38   1231   1107    162   44.28    53
  20 Lc0 0.22.0 J13B.2-188             :   3404      7    4500    3068.0   68.18   1955   2226    319   49.47    84
  21 Lc0 LS-LS 12.1                    :   3398      9    3000    2056.0   68.53   1299   1514    187   50.47    52
  22 Lc0k 0.23.2 384x30-T40-2036       :   3398      7    4250    2905.0   68.35   1923   1964    363   46.21    98
  23 Fat Fritz 1.0                     :   3385      9    3000    2012.0   67.07   1311   1402    287   46.73    83
  24 Lc0 0.22.0 32930                  :   3379      7    4500    2942.0   65.38   1821   2242    437   49.82    66
  25 Lc0 0.22.0 LS 12 preview          :   3377     10    2750    1876.0   68.22   1258   1236    256   44.95    56
  26 Stockfish 10                      :   3376      8    2750    1337.5   48.64    428   1819    503   66.15    72
  27 Lc0 0.23.2 61950                  :   3372      7    3750    2482.0   66.19   1550   1864    336   49.71    62
  28 Lc0 0.23.2 58613+                 :   3371     10    2500    1750.0   70.00   1200   1100    200   44.00    71
  29 Lc0 0.22.0 384x30-T40-1573        :   3367      9    2750    1846.0   67.13   1220   1252    278   45.53    96
  30 Lc0 0.25.1 3972_20k_TCEC          :   3355      9    2500    1646.5   65.86    944   1405    151   56.20    93
  31 Komodo 14                         :   3346      7    3750    1466.5   39.11    293   2347   1110   62.59    91
  32 Lc0 0.22.0 11260                  :   3338      9    2750    1755.5   63.84   1044   1423    283   51.75    99
  33 Houdini 6.03                      :   3326      5    7000    2736.0   39.09    657   4158   2185   59.40    70
  34 Komodo 13.3                       :   3324      7    3750    1379.0   36.77    294   2170   1286   57.87   100
  35 Komodo 13.2.5                     :   3310      6    4500    1749.5   38.88    503   2493   1504   55.40   100
  36 Scorpio 3.07 NN-ens-net-20x256    :   3280      9    2500    1466.5   58.66    855   1223    422   48.92   100
  37 Ethereal 12.00                    :   3223      7    5000    1244.5   24.89    207   2075   2718   41.50    99
  38 Xiphos 0.6                        :   3212      6    7000    1787.0   25.53    269   3036   3695   43.37    79
  39 Ethereal 11.75                    :   3208      8    4500    1196.0   26.58    210   1972   2318   43.82    94
  40 Fire 7.1                          :   3200      6    7000    1702.0   24.31    325   2754   3921   39.34   100
  41 RofChade 2.3                      :   3165      9    4500     866.5   19.26    108   1517   2875   33.71   100
  42 Booot 6.4                         :   3140      8    5000     861.0   17.22     96   1530   3374   30.60    78
  43 Laser 1.7                         :   3135      7    7000    1283.5   18.34    193   2181   4626   31.16    98
  44 Andscacs 0.95                     :   3124      7    7000    1221.0   17.44    200   2042   4758   29.17   100
  45 Deep Shredder 13                  :   3102      8    5000     818.0   16.36    114   1408   3478   28.16    84
  46 Booot 6.3.1                       :   3096      9    4750     767.0   16.15    118   1298   3334   27.33   ---

White advantage = 60.18 +/- 0.84
Draw rate (equal opponents) = 66.30 % +/- 0.23

Games        : 90.500 (finished)

White Wins   : 30.321 (33.5 %)
Black Wins   : 18.656 (20.6 %)
Draws        : 41.523 (45.9 %)

White Score  : 56.4 %
Black Score  : 43.6 %
http://www.fastgm.de

Alayan
Posts: 300
Joined: Tue Nov 19, 2019 7:48 pm
Full name: Alayan Feh

Re: FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Post by Alayan » Tue Jun 30, 2020 10:06 pm

Very weak performance.

The performance against SF11 wasn't that terrible, but the performance against weaker AB engines was not competitive at all. Less than 70% score against Ethereal 12, Xiphos 0.6 and Fire 7.1.

The big net isn't adapted to this fast of a TC, but this seems extreme.

fastgm
Posts: 552
Joined: Mon Aug 19, 2013 4:57 pm
Contact:

Re: FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Post by fastgm » Tue Jun 30, 2020 10:56 pm

Alayan wrote:
Tue Jun 30, 2020 10:06 pm
Very weak performance.
Yes, indeed!
75 Elo below the best 384x30 net, Lc0 0.24.1 384x30-T60-3010.
Plus a high draw rate, 56.2 %. Other nets have 10 % less!

fastgm
Posts: 552
Joined: Mon Aug 19, 2013 4:57 pm
Contact:

Re: FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Post by fastgm » Wed Jul 01, 2020 1:34 pm

The test with Lc0 0.25.1 3972_20k_TCEC must be repeated, because I missed an important parameter when calling the lc0.exe.
Three of the six mhl parameters are hidden parameters. These must be made visible and adjustable by the lc0.exe parameter "--show-hidden".

Many thanks to Stefan Pohl for his competent knowledge and help!

Alayan
Posts: 300
Joined: Tue Nov 19, 2019 7:48 pm
Full name: Alayan Feh

Re: FGRL - 16 Cores vs RTX 2070 rating list - 30. Lc0 0.25.1 3972_20k_TCEC (TCEC Superfinal S18)

Post by Alayan » Wed Jul 01, 2020 2:39 pm

We'll see if it makes any difference. With small sample size, it might be hard to tell if they end up close enough.

Both results should stay in the list afterwards, just with indications showing the used settings like we already have at the bottom of the page for several nets.

Post Reply