Chess2u
Would you like to react to this message? Create an account in a few clicks or log in to continue.

Chess2uLog in

stockfish of the neural network

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
ChessFan1 wrote:
SF-PB-100122_x64
Author: Michael Chaly
Date: Mon Jan 10 19:35:53 2022 +0100
Timestamp: 1641839753

Adjust pruning constants

Elo progress -> +30.51 https://github.com/glinscott/fishtest/wiki/Regression-Tests

Engine https://pixeldrain.com/u/KCB3bqaP


Some data errors in the file, can U pls upload again?

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Triton6654 wrote:
ChessFan1 wrote:
SF-PB-100122_x64
Author: Michael Chaly
Date: Mon Jan 10 19:35:53 2022 +0100
Timestamp: 1641839753

Adjust pruning constants

Elo progress -> +30.51 https://github.com/glinscott/fishtest/wiki/Regression-Tests

Engine https://pixeldrain.com/u/KCB3bqaP


Some data errors in the file, can U pls upload again?

Hmm, for some reason SSE3 engine is broken, just delete it. Here is link just for engines (without net) https://pixeldrain.com/u/tFCLxWqe

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Author: Michael Chaly
Date: Sat Jan 29 08:58:12 2022 +0100
Timestamp: 1643443092

https://abrok.eu/stockfish/

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
@lisandr Welcome back!

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
SF-PB-290122_x64, Updated to "Do stats updates after LMR for captures".

Engines https://pixeldrain.com/u/rXGg8PZ3

Network https://tests.stockfishchess.org/api/nn/nn-ac07bd334b62.nnue

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Author: Michael Chaly
Date: Sat Feb 5 07:40:21 2022 +0100
Timestamp: 1644043221

Author: Michael Chaly
Date: Sat Feb 5 07:38:30 2022 +0100
Timestamp: 1644043110

Author: ppigazzini
Date: Sat Feb 5 07:33:34 2022 +0100
Timestamp: 1644042814

Author: Michael Chaly
Date: Sat Feb 5 07:30:09 2022 +0100
Timestamp: 1644042609

https://abrok.eu/stockfish/

descriptionstockfish of the neural network - Page 16 Emptyhttps://abrok.eu/stockfish/

more_horiz
Author: Michael Chaly
Date: Mon Feb 7 17:30:35 2022 +0100
Timestamp: 1644251435

Do less depth reduction in null move pruning for complex positions

This patch makes us reduce less depth in null move pruning if complexity is high enough.
Thus, null move pruning now depends in two distinct ways on complexity,
while being the only search heuristic that exploits complexity so far.

passed STC
https://tests.stockfishchess.org/tests/view/61fde60fd508ec6a1c9f7754
LLR: 2.94 (-2.94,2.94) <0.00,2.50>
Total: 170000 W: 45555 L: 45027 D: 79418 Elo +1.08
Ptnml(0-2): 760, 19352, 44359, 19658, 871

passed LTC
https://tests.stockfishchess.org/tests/view/61fe91febf46cb834cbd5c90
LLR: 2.96 (-2.94,2.94) <0.50,3.00>
Total: 145272 W: 39182 L: 38651 D: 67439 Elo +1.27
Ptnml(0-2): 127, 14864, 42157, 15327, 161

closes https://github.com/official-stockfish/Stockfish/pull/3923

bench 4461945

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
SF-PB-070222_x64, Updated to "Do less depth reduction in null move pruning for complex positions".

Engines https://pixeldrain.com/u/Cr5HHuaV

Network https://tests.stockfishchess.org/api/nn/nn-ac07bd334b62.nnue

bench 4461945

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Nice thank you

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Author: Michael Chaly
Date: Wed Feb 9 17:17:00 2022 +0100
Timestamp: 1644423420

Big search tuning

Most credits for this patch should go to @candirufish.
Based on his big search tuning (1M games at 20+0.1s)

https://tests.stockfishchess.org/tests/view/61fc7a6ed508ec6a1c9f4b7d

with some hand polishing on top of it, which includes :

a) correcting trend sigmoid - for some reason original tuning resulted in it being negative. This heuristic was proven to be worth some elo for years so reversing it sign is probably some random artefact;
b) remove changes to continuation history based pruning - this heuristic historically was really good at providing green STCs and then failing at LTC miserably if we tried to make it more strict, original tuning was done at short time control and thus it became more strict - which doesn't scale to longer time controls;
c) remove changes to improvement - not really indended :).

passed STC
https://tests.stockfishchess.org/tests/view/6203526e88ae2c84271c2ee2
LLR: 2.94 (-2.94,2.94) <0.00,2.50>
Total: 16840 W: 4604 L: 4363 D: 7873 Elo +4.97
Ptnml(0-2): 82, 1780, 4449, 2033, 76

passed LTC
https://tests.stockfishchess.org/tests/view/620376e888ae2c84271c35d4
LLR: 2.96 (-2.94,2.94) <0.50,3.00>
Total: 17232 W: 4771 L: 4542 D: 7919 Elo +4.62
Ptnml(0-2): 14, 1655, 5048, 1886, 13

closes https://github.com/official-stockfish/Stockfish/pull/3926

bench 5030992

https://abrok.eu/stockfish/

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Strong patch sf

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue.

+10 ELO in two days...

https://abrok.eu/stockfish/

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Waldemar wrote:
Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue.

+10 ELO in two days...

https://abrok.eu/stockfish/


friends I believe that's not how it works, the number  of tested games is just too small compared to previous tests that is why the number of gain looks big.

Compare previous number of tested games:
Date: Sat Feb 5 07:40:21 2022 +0100
Total: 226120 W: 61106 L: 60472 D: 104542 Elo +0.97

With the most recent one:
Date: Thu Feb 10 19:54:31 2022 +0100
Total: 16952 W: 4775 L: 4521 D: 7656 Elo +5.21

We will have to wait for more tests in the future to conlcude the true elo gain!

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
in fact according to the most recent tests from NCM
the most recent update is slightly weaker that the on Feb 05
stockfish of the neural network - Page 16 Screen10

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
SF-PB-100222_x64, Updated to "Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue."

Engines https://pixeldrain.com/u/WsV4FB6N

Network https://tests.stockfishchess.org/api/nn/nn-6877cd24400e.nnue

Bench: 4919707

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
polo.toprak wrote:
in fact according to the most recent tests from NCM
the most recent update is slightly weaker that the on Feb 05
stockfish of the neural network - Page 16 Screen10


Excuse me, I don't seem to understand anything well...
Can you explain to me, how the derivative from 07022022 is weaker than the derivative from 05022022, if it has:

the highest no. of games won: 17479
the smallest no. of lost games: 28
the smallest no. draw game: 2493

Build Total Wins Losses Draws Elo
20220209-1617 20000 17364 35 2601 +458.15 +/- 6.66
20220207-1630 20000 17479 28 2493 +466.84 +/- 6.81
20220205-0640 20000 17360 30 2610 +458.22 +/- 6.65
20220205-0638 20000 17360 29 2611 +458.29 +/- 6.65
20220205-0633 20000 17363 44 2593 +457.45 +/- 6.68
20220205-0630 20000 17328 41 2631 +455.24 +/- 6.63

Thank you in advance.

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
ZamChess wrote:
polo.toprak wrote:
in fact according to the most recent tests from NCM
the most recent update is slightly weaker that the on Feb 05
stockfish of the neural network - Page 16 Screen10


Excuse me, I don't seem to understand anything well...
Can you explain to me, how the derivative from 07022022 is weaker than the derivative from 05022022, if it has:

the highest no. of games won: 17479
the smallest no. of lost games: 28
the smallest no. draw game: 2493

...



He was talking about "the most recent update" -> Feb 09

Cheers

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
SF Regression/Progression Tests
progression test of "Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue." of Feb 10th vs SF14.

ELO: 34.88 +-1.0 (95%) LOS: 100.0%

Total: 60000 W: 7532 L: 1528 D: 50940
Ptnml(0-2): 13, 884, 22430, 6432, 241

60000 @ 60+0.6 th 1

Source https://tests.stockfishchess.org/tests/view/620562ffd71106ed12a449a6

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Sf is really strong

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Let's begin winning of draws. First sf need solve all in 2030 I think so

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Edit

Last edited by deeds on Fri Jun 10, 2022 6:48 pm; edited 1 time in total

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Author: Michael Chaly
Date: Sun Feb 13 01:05:27 2022 +0100
Timestamp: 1644710727

Big search tuning (version 2)

One more tuning - this one includes newly introduced heuristics and
some other parameters that were not included in previous one. Result
of 400k games at 20+0.2 "as is". Tuning is continuing since there is
probably a lot more elo to gain.

STC:
https://tests.stockfishchess.org/tests/view/620782edd71106ed12a497d1
LLR: 2.99 (-2.94,2.94) <0.00,2.50>
Total: 38504 W: 10260 L: 9978 D: 18266 Elo +2.54
Ptnml(0-2): 142, 4249, 10230, 4447, 184

LTC:
https://tests.stockfishchess.org/tests/view/6207a243d71106ed12a49d07
LLR: 2.94 (-2.94,2.94) <0.50,3.00>
Total: 25176 W: 6793 L: 6546 D: 11837 Elo +3.41
Ptnml(0-2): 20, 2472, 7360, 2713, 23

closes https://github.com/official-stockfish/Stockfish/pull/3931

Bench: 4784796

https://abrok.eu/stockfish/

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
ChessFan1 wrote:
SF-PB-100222_x64, Updated to "Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue."

Engines https://pixeldrain.com/u/WsV4FB6N

Network https://tests.stockfishchess.org/api/nn/nn-6877cd24400e.nnue

Bench: 4919707


To my surprise, the 070222 version is still stronger, seem to be the SFNNv4 net did not benefit Stockfish PB.

https://pixeldrain.com/u/MZCs2PY5

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Triton6654 wrote:
ChessFan1 wrote:
SF-PB-100222_x64, Updated to "Update architecture to "SFNNv4". Update network to nn-6877cd24400e.nnue."

Engines https://pixeldrain.com/u/WsV4FB6N

Network https://tests.stockfishchess.org/api/nn/nn-6877cd24400e.nnue

Bench: 4919707


To my surprise, the 070222 version is still stronger, seem to be the SFNNv4 net did not benefit Stockfish PB.

https://pixeldrain.com/u/MZCs2PY5


How did you understand?
Have you run tests?

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Triton6654 wrote:

To my surprise, the 070222 version is still stronger, seem to be the SFNNv4 net did not benefit Stockfish PB.

https://pixeldrain.com/u/MZCs2PY5

Not my fault, all the complains to Stockfish team. bigsmile  Btw, in Stefan's Pohl tests SF 100222 is showing regression at TC: 3min+1sec (-3 elo).

Code:

     Program                    Elo    +    -   Games   Score   Av.Op.  Draws

   1 Stockfish 220205 avx2    : 3804    8    8  7000    73.3%   3622   52.7%
   2 Stockfish 220210 avx2    : 3801    8    8  7000    73.0%   3622   53.6%  < last tested
   3 Stockfish 14.1 211028    : 3780    7    7  8000    71.7%   3613   56.0%

https://www.sp-cc.de/index.htm

descriptionstockfish of the neural network - Page 16 EmptyRe: stockfish of the neural network

more_horiz
Permissions in this forum:
You cannot reply to topics in this forum