FGRL - 16 Cores rating list - 4. Lc0 0.24.1 LS 13

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Dann Corbit, Harvey Williamson

fastgm
Posts: 818
Joined: Mon Aug 19, 2013 6:57 pm

FGRL - 16 Cores rating list - 4. Lc0 0.24.1 LS 13

Post by fastgm »

New: 4. Lc0 0.24.1 LS 13

Code: Select all

   # Engine                            :    Elo  Error   Games    Points    (%)      W      D      L     D(%)   CFS
 -------------------------------------------------------------------------------------------------------------------
   1 Lc0 0.24.1 LS 14.2                :   3470     11    2750    2176.5   79.15   1675   1003     72   36.47    92
   2 Lc0 0.24.0 LS 14                  :   3459      9    3500    2677.0   76.49   1964   1426    110   40.74    92
   3 Lc0k 0.23.2 256x20-T40-1541       :   3450      9    3750    2853.0   76.08   2101   1504    145   40.11    53
   4 Lc0 0.24.1 LS 13                  :   3449     11    2500    1907.5   76.30   1401   1013     86   40.52    99
   5 Fat Fritz 1.1                     :   3433      9    3500    2661.5   76.04   1979   1365    156   39.00    69
   6 Lc0 0.24.1 384x30-T60-3010        :   3429     10    3000    2290.5   76.35   1673   1235     92   41.17    72
   7 Lc0 0.22.0 T40B.4-160             :   3425      8    4000    2906.0   72.65   2034   1744    222   43.60    55
   8 Stockfish 11 C0                   :   3425      7    3750    1911.0   50.96    520   2782    448   74.19    73
   9 Lc0 0.22.0 42850                  :   3421      8    4000    2889.5   72.24   2029   1721    250   43.02    65
  10 Stockfish 11                      :   3419      7    4000    2157.5   53.94    777   2761    462   69.03    99
  11 Lc0k 0.23.2 62172                 :   3404     10    2500    1827.0   73.08   1256   1142    102   45.68    50
  12 Lc0 LS-LS 12.2                    :   3404      8    3750    2680.5   71.48   1809   1743    198   46.48    78
  13 Lc0 0.22.0 J13B.2-188             :   3400      7    4000    2797.5   69.94   1856   1883    261   47.08    70
  14 Lc0 LS-LS 12.1                    :   3397      9    3000    2056.0   68.53   1299   1514    187   50.47    54
  15 Lc0k 0.23.2 384x30-T40-2036       :   3396      8    3750    2649.0   70.64   1837   1624    289   43.31    98
  16 Fat Fritz 1.0                     :   3383      9    3000    2012.0   67.07   1311   1402    287   46.73    82
  17 Lc0 0.22.0 32930                  :   3378      8    4000    2700.5   67.51   1744   1913    343   47.83    67
  18 Lc0 0.22.0 LS 12 preview          :   3375      9    2750    1876.0   68.22   1258   1236    256   44.95    57
  19 Stockfish 10                      :   3374      8    2750    1337.5   48.64    428   1819    503   66.15    67
  20 Lc0 0.23.2 61950                  :   3371      8    3750    2482.0   66.19   1550   1864    336   49.71    64
  21 Lc0 0.23.2 58613+                 :   3369     10    2500    1750.0   70.00   1200   1100    200   44.00    71
  22 Lc0 0.22.0 384x30-T40-1573        :   3365      9    2750    1846.0   67.13   1220   1252    278   45.53   100
  23 Lc0 0.22.0 11260                  :   3337      9    2750    1755.5   63.84   1044   1423    283   51.75    98
  24 Houdini 6.03                      :   3325      6    5250    2098.0   39.96    535   3126   1589   59.54    60
  25 Komodo 13.3                       :   3324      8    3250    1197.5   36.85    268   1859   1123   57.20   100
  26 Komodo 13.2.5                     :   3309      6    4500    1749.5   38.88    503   2493   1504   55.40   100
  27 Scorpio 3.07 NN-ens-net-20x256    :   3278      9    2500    1466.5   58.66    855   1223    422   48.92   100
  28 Ethereal 12.00                    :   3226      9    3250     823.5   25.34    150   1347   1753   41.45   100
  29 Xiphos 0.6                        :   3208      7    5250    1360.0   25.90    218   2284   2748   43.50    63
  30 Ethereal 11.75                    :   3206      7    4500    1196.0   26.58    210   1972   2318   43.82    89
  31 Fire 7.1                          :   3200      7    5250    1316.0   25.07    260   2112   2878   40.23   100
  32 RofChade 2.3                      :   3163     11    2750     527.0   19.16     75    904   1771   32.87   100
  33 Booot 6.4                         :   3138     11    3250     559.5   17.22     64    991   2195   30.49    80
  34 Laser 1.7                         :   3132      8    5250     982.0   18.70    164   1636   3450   31.16    95
  35 Andscacs 0.95                     :   3123      8    5250     940.0   17.90    171   1538   3541   29.30   100
  36 Deep Shredder 13                  :   3101      9    5000     818.0   16.36    114   1408   3478   28.16    83
  37 Booot 6.3.1                       :   3094      9    4750     767.0   16.15    118   1298   3334   27.33   ---

White advantage = 61.23 +/- 0.98
Draw rate (equal opponents) = 64.58 % +/- 0.27

Games        : 68.000 (finished)

White Wins   : 23.254 (34.2 %)
Black Wins   : 14.416 (21.2 %)
Draws        : 30.330 (44.6 %)

White Score  : 56.5 %
Black Score  : 43.5 %
All 68.000 games: http://www.fastgm.de/schach/16cores.7z

Next match: Stockfish 200502 Contempt 0

http://www.fastgm.de
User avatar
pohl4711
Posts: 2391
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: FGRL - 16 Cores rating list - 4. Lc0 0.24.1 LS 13

Post by pohl4711 »

Perhaps time for new T60 testrun? Your latest T60 testrun was 62172.
In my MEA-testings Nets 62100 and Net 62200 had a scorerate of 0.649 and 0.650. The best T60 net so far (63305) had a scorerate of 0.657. As I wrote on my website:
"I made some comparsions of some lc0-results on discord and in the FGRL-ratinglist. And it seems, that an increasing scorerate of +0.001 is around +2 Elo gain...but only with nets of the same size."
If this is correct, I would predict an Elo-gain of around +14 Elo for 63305, compared to 62172. Not very much, but it would lift T60 on a level with Stockfish 11 in your ratinglist.

Keep up the brillant work!

Code: Select all

1 node (the onboard-position) calculated, only (!) 
(Thinking-time: 1ms and Option Slowmover=0 makes lc0 calculate only 
1 node and then move instantly, no matter, how long the calculation
of 1 node takes)

Latest T60 net tested so far: 63410

Size: 24x320 (sorted by best Score):
Engine                           : Top1  Top1Rate  Score  ScoreRate

lc0 0.24.1 63305 (24x320)        : 21816   0.436  328355    0.657 <---
lc0 0.24.1 63330 (24x320)        : 21839   0.437  328124    0.656
lc0 0.24.1 63310 (24x320)        : 21828   0.437  327931    0.656
lc0 0.24.1 63350 (24x320)        : 21801   0.436  327754    0.656
lc0 0.24.1 63300 (24x320)        : 21776   0.436  327723    0.655
lc0 0.24.1 63320 (24x320)        : 21809   0.436  327601    0.655
lc0 0.24.1 63250 (24x320)        : 21839   0.437  327549    0.655
lc0 0.24.1 63275 (24x320)        : 21823   0.436  327515    0.655
lc0 0.24.1 63340 (24x320)        : 21781   0.436  327301    0.655
lc0 0.24.1 63370 (24x320)        : 21725   0.434  327225    0.654
lc0 0.24.1 63380 (24x320)        : 21757   0.435  327168    0.654
lc0 0.24.1 63200 (24x320)        : 21803   0.436  327162    0.654
lc0 0.24.1 63225 (24x320)        : 21735   0.435  326991    0.654
lc0 0.24.1 63100 (24x320)        : 21682   0.434  326936    0.654
lc0 0.24.1 63000 (24x320)        : 21761   0.435  326868    0.654
lc0 0.24.1 63390 (24x320)        : 21736   0.435  326746    0.653
lc0 0.24.1 62800 (24x320)        : 21737   0.435  326730    0.653
lc0 0.24.1 63360 (24x320)        : 21737   0.435  326716    0.653
lc0 0.24.1 63410 (24x320)        : 21730   0.435  326458    0.653
lc0 0.24.1 62400 (24x320)        : 21713   0.434  326424    0.653
lc0 0.24.1 62900 (24x320)        : 21740   0.435  326390    0.653
lc0 0.24.1 63400 (24x320)        : 21724   0.434  326265    0.653
lc0 0.24.1 62500 (24x320)        : 21633   0.433  325990    0.652
lc0 0.24.1 62700 (24x320)        : 21661   0.433  325977    0.652
lc0 0.24.1 62300 (24x320)        : 21672   0.433  325838    0.652
lc0 0.24.1 J13B.2-136 (24x320)   : 21481   0.430  325236    0.650
lc0 0.24.1 62100 (24x320)        : 21520   0.430  324995    0.650 <---
lc0 0.24.1 62200 (24x320)        : 21528   0.431  324637    0.649 <---
lc0 0.24.1 62000 (24x320)        : 21464   0.429  324211    0.648
lc0 0.24.1 62600 (24x320)        : 21502   0.430  324157    0.648
fastgm
Posts: 818
Joined: Mon Aug 19, 2013 6:57 pm

Re: FGRL - 16 Cores rating list - 4. Lc0 0.24.1 LS 13

Post by fastgm »

Thanks Stefan, that's a good tip!

I will test a new T60 net after the Stockfish match.
It would be nice if the MEA-test could predict this so accurately.

Andreas
User avatar
pohl4711
Posts: 2391
Joined: Sat Sep 03, 2011 7:25 am
Location: Berlin, Germany
Full name: Stefan Pohl

Re: FGRL - 16 Cores rating list - 4. Lc0 0.24.1 LS 13

Post by pohl4711 »

fastgm wrote: Tue May 05, 2020 8:52 pm Thanks Stefan, that's a good tip!

I will test a new T60 net after the Stockfish match.
It would be nice if the MEA-test could predict this so accurately.

Andreas
Net 63305 is the best so far in my MEA testings. All newer T60 are weaker, especially the latest (63430, 63420)...