Sergio Vieri second net is out

[MikeB]

The last (2210) finally 59%, +63 elo above the official stockfish 11.

Code: Select all

SF-NNUE (popcount) vs Stockfish 11
henk-2706     54.7%   2020-07-19
sergio-1432   56.1%   2020-07-21 15:56
sergio-1907   58.9%   2020-07-21 19:14
sergio-2323   58.4%   2020-07-21 23:23
sergio-0359   58.8%   2020-07-22 03:59
sergio-0944   58.9%   2020-07-22 09:44
sergio-1153   58.3%   2020-07-22 11:53
sergio-1807   58.0%   2020-07-22 18:07
sergio-2210   59.0%   2020-07-22 22:10

I get a strong regression for 20200723-1134.bin

Same here, from 59.0% to 56.5%, a loss of almost 20 elo.

I have found in the past , when evaluating nets , when you see 'experimental' - avoid.

I have tested the other nets and it my belief 2141 is the current strongest I have tested, yomv , although the elo gaining is slowing which it be expected.

Not every net is a gainer - 1844 was experimental and failed. (correction)

Code: Select all

 4000 game(s) loaded
Rank Name    Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------
   1 sf2141   3503   0.0    7    7  4000 2037.5  50.9 1002  927 2071  25.1  51.8  3497
   2 sf1844   3497   6.7    7    7  4000 1962.5  49.1  927 1002 2071  23.2  51.8  3503
---------------------------------------------------------------------------------------------------------
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
        sf sf
sf2141     96
sf1844   3

4000 game(s) loaded

the latest 1134 will be tested next

[jmartus]

Its really hard to understand the moves there making

k1br1b1r/1pq4p/3p1pn1/pPp1n1p1/4P3/N1P1B2P/P1Q2PP1/3RRBNK w - - 0 29 b6! why sacrifice the pawn like that? Sergio722 2115 Net

[MikeB]

so far - 2141 (my current leader, yomv) - vs the challenger 1134 released today

Code: Select all

tc/base+inc: 60+0.60
games planned: 4000

Current date : time (EDST)
Date: 07/23/20 : 13:20:42

176 game(s) loaded
Rank Name    Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------
   1 sf1134   3508   0.0   33   33   176   92.0  52.3   43   35   98  24.4  55.7  3492
   2 sf2141   3492  15.8   33   33   176   84.0  47.7   35   43   98  19.9  55.7  3508
---------------------------------------------------------------------------------------------------------
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
        sf sf
sf1134     82
sf2141  17

also , I'm in the process of updating my scripts ( based Bob Hyatts scripts he used with crafty) to work with cutechess-cli ( finally) - that will produce the above output periodically when you run the cutechess client ( not the GUI version) runs one script with two other called scripts running in the background - it totally removes the need for the engine.json file which is one PIA - you set the variable once , not multiple times as as you do with the json file

Some features, it will automatically determine the best interval to output based on your personal desires, , number of engines, number of rounds time control, etc ( based plus increment), etc

- if there is interest - please let me know.

[zullil]

Its really hard to understand the moves there making

[d]k1br1b1r/1pq4p/3p1pn1/pPp1n1p1/4P3/N1P1B2P/P1Q2PP1/3RRBNK w - - 0 29 b6! why sacrifice the pawn like that? Sergio722 2115 Net

[Raphexon]

Its really hard to understand the moves there making

k1br1b1r/1pq4p/3p1pn1/pPp1n1p1/4P3/N1P1B2P/P1Q2PP1/3RRBNK w - - 0 29 b6! why sacrifice the pawn like that? Sergio722 2115 Net

It thinks the position is lost and that will minimize the loss.
SF_dev_20 april also picks b6, and thinks the position is even more lost.
In fact Lc0 with 128x10-t60-2-2010.pb net picks it too, also thinks the position is lost.

[zullil]

Its really hard to understand the moves there making

k1br1b1r/1pq4p/3p1pn1/pPp1n1p1/4P3/N1P1B2P/P1Q2PP1/3RRBNK w - - 0 29 b6! why sacrifice the pawn like that? Sergio722 2115 Net

It thinks the position is lost and that will minimize the loss.
SF_dev_20 april also picks b6, and thinks the position is even more lost.
In fact Lc0 with 128x10-t60-2-2010.pb net picks it too, also thinks the position is lost.

??? The position is lost for whom?

info depth 48 seldepth 88 multipv 1 score cp 311 nodes 29341363330 nps 24996731 hashfull 997 tbhits 0 time 1173808 pv b5b6 c7b6 d1b1 b6c7 b1b2 c8d7 a3b5 d7b5 f1b5 d8b8 c2a4 f8e7 g1e2 h8c8 e2d4 b7b6 d4e6 c7b7 e1b1 a8a7 b5e2 e7d8 g2g3 b7d7 a4d7 e5d7 b2d2 d7e5 d2d6 d8e7 d6d5 c8c6 e6g7 c6d6 d5d6 e7d6 b1d1 e5f7 e2c4 f7e5 c4d5 c5c4 g7e6 g6e7 f2f4 g5f4 g3f4 e5d3 d5c4 d3b2 d1d4 b2c4 d4c4 a7b7 c4d4 b7c6 d4d1 d6c5 e6d4 c5d4 c3d4 f6f5 d4d5 c6d7 d5d6 e7g6 e4f5 g6h4

Seems the obvious point of b6 is to open the b-file and the b5 square.

[Raphexon]

Its really hard to understand the moves there making

k1br1b1r/1pq4p/3p1pn1/pPp1n1p1/4P3/N1P1B2P/P1Q2PP1/3RRBNK w - - 0 29 b6! why sacrifice the pawn like that? Sergio722 2115 Net

It thinks the position is lost and that will minimize the loss.
SF_dev_20 april also picks b6, and thinks the position is even more lost.
In fact Lc0 with 128x10-t60-2-2010.pb net picks it too, also thinks the position is lost.

??? The position is lost for whom?

info depth 48 seldepth 88 multipv 1 score cp 311 nodes 29341363330 nps 24996731 hashfull 997 tbhits 0 time 1173808 pv b5b6 c7b6 d1b1 b6c7 b1b2 c8d7 a3b5 d7b5 f1b5 d8b8 c2a4 f8e7 g1e2 h8c8 e2d4 b7b6 d4e6 c7b7 e1b1 a8a7 b5e2 e7d8 g2g3 b7d7 a4d7 e5d7 b2d2 d7e5 d2d6 d8e7 d6d5 c8c6 e6g7 c6d6 d5d6 e7d6 b1d1 e5f7 e2c4 f7e5 c4d5 c5c4 g7e6 g6e7 f2f4 g5f4 g3f4 e5d3 d5c4 d3b2 d1d4 b2c4 d4c4 a7b7 c4d4 b7c6 d4d1 d6c5 e6d4 c5d4 c3d4 f6f5 d4d5 c6d7 d5d6 e7g6 e4f5 g6h4

Seems the obvious point of b6 is to open the b-file and the b5 square.

Only did a check at shallow depth.
I can try a longer check too.

[MikeB]

so far - 2141 (my current leader, yomv) - vs the challenger 1134 released today

Code: Select all

tc/base+inc: 60+0.60
games planned: 4000

Current date : time (EDST)
Date: 07/23/20 : 13:20:42

176 game(s) loaded
Rank Name    Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------
   1 sf1134   3508   0.0   33   33   176   92.0  52.3   43   35   98  24.4  55.7  3492
   2 sf2141   3492  15.8   33   33   176   84.0  47.7   35   43   98  19.9  55.7  3508
---------------------------------------------------------------------------------------------------------
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
        sf sf
sf1134     82
sf2141  17

also , I'm in the process of updating my scripts ( based Bob Hyatts scripts he used with crafty) to work with cutechess-cli ( finally) - that will produce the above output periodically when you run the cutechess client ( not the GUI version) runs one script with two other called scripts running in the background - it totally removes the need for the engine.json file which is one PIA - you set the variable once , not multiple times as as you do with the json file

Some features, it will automatically determine the best interval to output based on your personal desires, , number of engines, number of rounds time control, etc ( based plus increment), etc

- if there is interest - please let me know.

latest , NN ,time stamped 1134 looks like a gainer ... current best in my book, yomv

Code: Select all

pgn file: c:/cluster.mfb/pgn/072320-2.pgn
tc/base+inc: 60+0.60
games planned: 4000

Current date : time (EDST)
Date: 07/23/20 : 17:24:12

4000 game(s) loaded
Rank Name    Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------
   1 sf1134   3503   0.0    7    7  4000 2031.0  50.8  947  885 2168  23.7  54.2  3497
   2 sf2141   3497   5.4    7    7  4000 1969.0  49.2  885  947 2168  22.1  54.2  3503
---------------------------------------------------------------------------------------------------------
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
        sf sf
sf1134     93
sf2141   6

[MMarco]

Seems the obvious point of b6 is to open the b-file and the b5 square.

Yes. Also note that if it would be black's turn, they would play b6 to close the queenside and avoid white's attack there. You can verify this by adding a null move. According to Bluefish-XI, black would get a slight edge if it could play b6 for free:

Code: Select all

[28] -0.47 b6 Cc4 h5 Fc1 Fh6 Ce3 Ce7 Fc4 g4 Cd5 Da7 Fxh6 Txh6 Cxe7 Dxe7 Fd5+ Rb8 Dd2 Tg6 Te3 Fe6 Fxe6 Dxe6 Dd5 Dxd5 Txd5 h4 Ce2 Cc4 Ted3 Te8 Cf4 Tg5 Txd6 Cxd6
d:28/45 t:75.90 n:223M nps:2946k h:94

[MikeB]

Net 20200723-1134.bin is still champion.
Net 20200724-0640.bin up next

Code: Select all

pgn file: c:/cluster.mfb/pgn/072320-3.pgn
tc/base+inc: 60+0.60
games planned: 4000

Current date : time (EDST)
Date: 07/23/20 : 23:27:18

4000 game(s) loaded
Rank Name    Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR
---------------------------------------------------------------------------------------------------------
   1 sf1134   3501   0.0    7    7  4000 2010.5  50.3  940  919 2141  23.5  53.5  3499
   2 sf0123   3499   1.8    7    7  4000 1989.5  49.7  919  940 2141  23.0  53.5  3501
---------------------------------------------------------------------------------------------------------
  Σ = total score, 1 point for win, 1/2 point for draw

LOS:
        sf sf
sf1134     69
sf0123  30

4000 game(s) loaded