LS-ratinglist: Stockfish 130623

[pohl4711]

The result of Stockfish 130623 is now online.

http://ls-ratinglist.beepworld.de/


Stefan

(Perhaps you have to clear your browsercache or reload the website)

[Vinvin]

Thanks for test !
That suggests +35 elo since SF 3 !!

Code: Select all

23 Stockfish 3 x64s        3019    5    5 11000   46%  3050   45% 
15 Stockfish 130519 x64s   3037    5    5 11000   49%  3048   45% 
13 Stockfish 130601 x64s   3046    4    5 13000   50%  3048   45% 
12 Stockfish 130623 x64s   3054    5    5 10000   50%  3052   46%

[pohl4711]

Thanks for test !
That suggests +35 elo since SF 3 !!

Yes, thats what I wrote in the headline of the Home-Screen of my website...+35 Elo to Stockfish 3.

[Vinvin]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

[pohl4711]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

IMO nobody needs a official release as long as the development-compiles work so fine (and faster as the official JA-compiles)...

Best - Stefan

[Laskos]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

This won't be +35 or +30 points, it's at bullet control. +15 is more reasonable to expect at 40/4 or 5+3.

[geots]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

This won't be +35 or +30 points, it's at bullet control. +15 is more reasonable to expect at 40/4 or 5+3.

+2

gts

[pohl4711]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

This won't be +35 or +30 points, it's at bullet control. +15 is more reasonable to expect at 40/4 or 5+3.

Interesting. Especially when you compare the Elo-difference between Stockfish 2.3.1 and Stockfish 3 in the IPON (+22 Elo) and the LS-ratinglist (+24 Elo)... And IPON is 5+3...

Stefan

[Ajedrecista]

Hello Stefan:

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

This won't be +35 or +30 points, it's at bullet control. +15 is more reasonable to expect at 40/4 or 5+3.

Interesting. Especially when you compare the Elo-difference between Stockfish 2.3.1 and Stockfish 3 in the IPON (+22 Elo) and the LS-ratinglist (+24 Elo)... And IPON is 5+3...

Stefan

Thanks for your effort.

I also think that +35 Elo at this point may be a little overestimated due to bullet TC. However, please take a look to this web:

Stockfish foreign engines regression test

TC is 2' + 0.05"/move. At this moment, error bars are not larger than ± 7 Elo. Combining the rating and error bars of SF 3 and the latest tested compile there (Include file attacks in 'major on pawn'), the rating difference is more less 2793 - 2774 ± sqrt[(6.5)² + 7²] ~+19 ± 9.55 Elo. The pool of the engines is not very wide but also the TC can not be considered bullet.

@Vincent: the date release of SF 3 was on April, 30th. I think that SF 3 do not want to flood us with official releases. Of course, there are tons of autocompiles at abrok.eu. From LS rating list, I can deduce that the rating difference including error bars is more less 3054 - 3019 ± sqrt(5² + 5²) = +35 ± 5*sqrt(2) ~ +35 ± 7.07 Elo; probably the true rating difference falls in the lower region of the interval due to bullet TC, but it is just a weird guess.

Regards from Spain.

Ajedrecista.

[gladius]

IMO, +30 Elo is enough to release a new official version of a freeware chess engine !

This won't be +35 or +30 points, it's at bullet control. +15 is more reasonable to expect at 40/4 or 5+3.

Interesting. Especially when you compare the Elo-difference between Stockfish 2.3.1 and Stockfish 3 in the IPON (+22 Elo) and the LS-ratinglist (+24 Elo)... And IPON is 5+3...

Stefan

I hope it holds up a longer TC as well . Who knows though. In self tests against SF 3 at 1min+0.05, the current development version has:

ELO: 35.68 +-2.2 (95%) LOS: 100.0%
Total: 38318 W: 9849 L: 5928 D: 22541

Whereas, SF 3 vs SF 2.3.1 had:

ELO: 30.86 +-2.9 (95%) LOS: 100.0%
Total: 20000 W: 4493 L: 2721 D: 12786