Stockfish 16.1 release candidate

[Disservin]

Todo:

• draft release notes

• finish progression tests:

  https://tests.stockfishchess.org/tests/view/65d666051d8e83c78bfddbd6
  https://tests.stockfishchess.org/tests/view/65d666051d8e83c78bfddbd8

• watch out for release blocking issues

• update WDL model

Bench: 1303971

[mstembera]

The elo gain from 16 to 16.1 will be higher than 15 to 16 was.

[dav1312]

The elo gain from 16 to 16.1 will be higher than 15 to 16 was.

SF 16 vs SF 15 https://tests.stockfishchess.org/tests/view/6494097ddc7002ce609c99b7

new_tc	60+0.6
threads	1
book	UHO_XXL_+0.90_+1.19.epd

Elo: 47.03 ± 1.3 (95%) LOS: 100.0%
Total: 60000 W: 20247 L: 12174 D: 27579
Ptnml(0-2): 26, 2938, 16102, 10805, 129
nElo: 103.71 ± 3.0 (95%) PairsRatio: 3.69

SF 16.1 vs SF 16 https://tests.stockfishchess.org/tests/view/65d666051d8e83c78bfddbd6

new_tc	60+0.6
threads	1
book	UHO_4060_v3.epd

Elo: 27.66 ± 1.6 (95%) LOS: 100.0%
Total: 41960 W: 12357 L: 9024 D: 20579
Ptnml(0-2): 47, 3353, 10930, 6520, 130
nElo: 56.39 ± 3.4 (95%) PairsRatio: 1.96

[mstembera]

@dav1312 I was comparing the currently running progression tests to the 1 thread progression chart at https://github.com/official-stockfish/Stockfish/wiki/Regression-Tests where it shows that 15 to 16 was less than 20 elo.

[PGG106]

Different books, 8moves vs uho, if you look at the column for the uho book in the page you sent you'll see this

Elo: [47.03] ± 1.3
Ptnml: 26, 2938, 16102, 10805, 129
nElo: 103.71 ±3.0
PairsRatio: 3.69

47 > 27

[dav1312]

@dav1312 I was comparing the currently running progression tests to the 1 thread progression chart at Wiki: Regression Tests () where it shows that 15 to 16 was less than 20 elo.

You can't compare completely different books. That is why I sent you the data and links directly, you are looking at the wrong test.

[mstembera]

Ok that makes sense. Thanks for pointing out the book difference. I think comparing the results from these various books that are actually not comparable on the same chart like this one https://docs.google.com/spreadsheets/d/e/2PACX-1vQqw86SXD_-zzP39DzfjBQ1eLBGyZMPyVLPuZDTY7zSNxBvxxj9CUXpd_AHRKy1aCpCCXGsznolmMVs/pubchart?oid=1631702142&format=image causes misinterpretation.

[robertnurnberg]

One thing to add on the TODO list: [] update WDL model. We can do this at the very last minute (to have the model as accurate as possible), as it will only change the 8 coefficients in the two polynomials.

[Disservin]

@robertnurnberg yes good point, I can merge that together with the merge of sf 16.1

[Disservin]

merged with e67cc97, thank you all!