Sergio Vieri second net is out

[peter]

The actual NNUE implementation is handcrafted and only has AVX2, SSE2 and SSE4.1 implementation. So no additional compiler switches would make any difference at all.

I see, so I guess sse4.1-popcnt would be best for me and maybe some other user too.

[MikeB]

new leader with nice a pickup, ymmv

Code: Select all

###                                              Summary                                              ###
#########################################################################################################

PGN File: c:/cluster.mfb/pgn/08070134.pgn
Time Control: Time Control-> base+inc: 25+0.500
Games: 8000
Threads: 1
Hash: 256

Current date : time (EDST)
Date: 08/07/20 : 05:50:58

Projected-> Time: 4h:22m:27s
     Run -> Time: 4h:16m:36s

8000 game(s) loaded
Rank Name          Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------

   1 SF-XIr4-1802   3503   0.0    4    4  8000 4065.0  50.8 1422 1292 5286  17.8  66.1  3497
   2 SF-XIr4-2257   3497   5.8    4    4  8000 3935.0  49.2 1292 1422 5286  16.1  66.1  3503
---------------------------------------------------------------------------------------------------------

  Δ = delta from the next higher rated opponent
  # = number of games played
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
              SF SF
SF-XIr4-1802     99
SF-XIr4-2257   0

#########################################################################################################
###                                                End                                                ###
#########################################################################################################


8000 game(s) loaded

[Laskos]

Can somebody reproduce 30 Elo gain shown in SV plot since net 0943 (after 7-8 nets)? My today's results are nowhere close to these 30 Elo points.

[mehmet123]

In my tests elo gain is less than 5 elo both 1 min +0 .5 sec test and 2 min + 0.5 sec time control test.

[Laskos]

In my tests elo gain is less than 5 elo both 1 min +0 .5 sec test and 2 min + 0.5 sec time control test.

Same here.

[mehmet123]

Self elo tests usually give very exaggerated results. In Lc0 tests some time x self elo and x + 100 self elo engines showed similar performances against Stockfish.

[Sylwy]

1056 - the best Sergio Vieri net in my tests.
Conditions:
-TC=4'+2"
-Hash=256 MB
-GUI: Arena 3.5.1
-Book: IM_4mvs.abk (8 plies book) for both engines
-default settings for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-6-men Syzygy bases for both engines.

[MikeB]

Some sha256sums:
6b92d1ba8fc5 *20200720-1017.bin
e0fff38dfef2 *20200721-0909.bin
3ec62799130c *20200721-1432.bin
4d49fe98f8db *20200721-1907.bin
6bc4b6918d98 *20200721-2323.bin
32c96cd1eab7 *20200722-0359.bin
982724123539 *20200722-0944.bin
4ef3a2936f1d *20200722-1517.bin
effc2d2dd3e9 *20200722-2141.bin
d400e1af2e62 *20200723-0511.bin
4d410ae3956e *20200723-1134.bin
d02d795e1f26 *20200723-1844.bin
c85ac04dcee6 *20200724-0123.bin
b26eb318dcdb *20200724-0640.bin
ee269e2416d1 *20200724-1240.bin
5f043ed6bafe *20200724-1732.bin
ae91ca26c034 *20200724-2343.bin
f4275f82a556 *20200724-2344.bin
f28006c860fa *20200725-0545.bin
221e4b0e6303 *20200725-1313.bin
db77c9597204 *20200725-2051.bin
c5a7cbea67cf *20200726-0504.bin
ab01fc825f70 *20200726-1135.bin
f2e65be88d87 *20200727-0332.bin
917a774c37b3 *20200727-0928.bin
c8283dc505f9 *20200727-1540.bin
ca7fa8535b4c *20200727-2151.bin
e0f0ba36086e *20200728-0207.bin
91fa62ee6392 *20200728-0633.bin
b6ba221f1d14 *20200728-1104.bin
4667dfcaebd3 *20200728-1442.bin
a1781570fc41 *20200728-1817.bin
6fe21ce09072 *20200728-2138.bin
27863d755b26 *20200729-0109.bin
7a0941bf031d *20200729-0335.bin
6990bf4e9d9b *20200729-0629.bin
4d3e0ad7b628 *20200729-0912.bin
e6809c8c0140 *20200729-1218.bin
d571727494a6 *20200729-1500.bin
1f4743cfb5b2 *20200729-1743.bin
0825bff03ced *20200729-1910.bin
d996351232b3 *20200729-2214.bin
fc1c4bc8a813 *20200730-1934.bin
2d0ed6ba293d *20200730-2333.bin
26172af8f8eb *20200731-0111.bin
263296d96a07 *20200731-0252.bin
3e471000a8ec *20200731-0429.bin
adb83d813a58 *20200731-0631.bin
fe0dd44cdda4 *20200731-0821.bin
e72e4f291fdc *20200731-1607.bin
97f742aaefcd *20200801-1515.bin
65ef0e0d5450 *20200802-2257.bin
5f472394322d *20200803-2108.bin
648f11ec7bbb *20200804-1817.bin
4e938d6bbccd *20200805-0633.bin
56fcb934b42f *20200805-1056.bin
43eccdcad08e *20200805-1512.bin
40bc1c93114a *20200806-0943.bin
5e0c7d0f5e65 *20200806-1403.bin
9931db908a9b *20200806-1802.bin
ae5315ef084e *20200807-1645.bin
9a2e6989cc1e *20200807-1950.bin
5570a3425229 *20200808-0049.bin
64a95ba7f2d0 *20200808-0418.bin
5a3983417e4f *20200808-0817.bin
79da3a6b905b *20200808-1351.bin
e8d87fb8f0e5 *20200809-0854.bin
947e6629ebdb *20200809-1443.bin
e2c48381ce68 *20200809-1951.bin

112bb1c8cdb5 *20200810-0744.bin current best per Sergio and SF development team ...
https://github.com/official-stockfish/S ... a08b46dc6b

4025858e956e *experimental-20200722-0835.bin
e99689f8a257 *experimental-20200722-1130.bin
b6ddc6a86194 *experimental-20200722-1621.bin
6647051bd14c *experimental-20200722-2115.bin
b7051daa0ae1 *experimental-20200723-1334.bin
bfd1ea660291 *experimental-20200724-0650.bin
dd6f24f3f306 *experimental-20200724-1215.bin
f4275f82a556 *experimental-20200724-2244.bin
ce4a24307967 *experimental-20200725-2237.bin
c157e0a5755b *nn.bin >> was this the very first one released ??? - I don't see match , but maybe I missed it...
4d410ae3956e *rapid.bin >> fake, same as 1134