Stockfish (20 kn/s) vs Fritz in Bahrain(3600 kn/s)

[mehmet1921]

Program Elo + - Games Score Av.Op. Draws
1 Stockfish 271119 64 BMI2 : 2969 74 72 90 71.7 % 2807 14.4 %
2 Fritz in Bahrain : 2807 72 74 90 28.3 % 2969 14.4 %
(Arena Gui, 5 minutes/game, core i7 9750h, Book2014 opening book, ponder off)

Individual statistics:

1 Stockfish 271119 64 BMI2 : 2969 90 (+ 58,= 13,- 19), 71.7 %

Fritz in Bahrain : 90 (+ 58,= 13,- 19), 71.7 %

2 Fritz in Bahrain : 2807 90 (+ 19,= 13,- 58), 28.3 %

Stockfish 271119 64 BMI2 : 90 (+ 19,= 13,- 58), 28.3 %

Speed of Stockfish 271119 :20 kN/s
Speed of Fritz in Bahrain :3600 kN/s

Games of the Match:
http://www.mediafire.com/file/zyxwupci2 ... 4.pgn/file

Fritz in Bahrain vs World Chess Champion Kramnik (2807 elo) match has ended with a 4-4 score at 2002.
Speed of Fritz in Bahrain was 2800 kn/s at this match.

The human elo of Stockfish with 20 kn/s is ~3000 according to this test

[Jouni]

How do You limit SF speed to 20 kn/s in i7 CPU? PGN shows something else.

[mehmet1921]

I set nodestime parameter 20 at Stockfish settings.
Pgn shows a higher kn/s but Stockfish use a little time. Total nodes of Stockfish is ~ 1/180 of Fritz in Bahrain.

[MikeB]

How do You limit SF speed to 20 kn/s in i7 CPU? PGN shows something else.

That is also a feature I added in the latest release - on my machine nps 20,000 is equal to nps level 35.

https://github.com/MichaelB7/Stockfish/releases/tag/XR7

Included in the release is Stockfish XR7 , which is functionally equivalent of stockfish , but with the bells and whistles of Honey. Each of the downloads include 10 engines , including Stockfish XR7.

https://github.com/MichaelB7/Stockfish/releases/tag/XR7

https://www.dropbox.com/s/s1edcna7vtnar ... M.jpg?dl=0

[Vinvin]

That reminds me the challenge : Crafty-Rebel 100:1 time handicap

http://www.rebel.nl/match.htm

[Raphexon]

Should have used increments.

Looking at a few PGNs it looked like SF made some time related blunders that could have been avoided with even 0.5 second increments.
Fritz Bahrain may have made those too.

[Uri Blass]

Should have used increments.

Looking at a few PGNs it looked like SF made some time related blunders that could have been avoided with even 0.5 second increments.
Fritz Bahrain may have made those too.

With different conditions you get different results.
There are no right or wrong conditions.

[Raphexon]

Should have used increments.

Looking at a few PGNs it looked like SF made some time related blunders that could have been avoided with even 0.5 second increments.
Fritz Bahrain may have made those too.

With different conditions you get different results.
There are no right or wrong conditions.

There isn't, but using classical time controls or increments is more representative for what's normally played.
Using something like x/game is fine but needs aggressive adjudication if you don't want results that wouldn't have happened at other TCs.

It's just kinda jarring to see any engine throw away a sure win for a sure loss because it was out of time and reacted instantly, then again that's what time management is for. And time management is an art.

But using classical time control is really important if you want to make the claim of the opening post: "The human elo of Stockfish with 20 kn/s is ~3000 according to this test"

[MikeB]

Should have used increments.

Looking at a few PGNs it looked like SF made some time related blunders that could have been avoided with even 0.5 second increments.
Fritz Bahrain may have made those too.

With different conditions you get different results.
There are no right or wrong conditions.

There isn't, but using classical time controls or increments is more representative for what's normally played.
Using something like x/game is fine but needs aggressive adjudication if you don't want results that wouldn't have happened at other TCs.

It's just kinda jarring to see any engine throw away a sure win for a sure loss because it was out of time and reacted instantly, then again that's what time management is for. And time management is an art.

But using classical time control is really important if you want to make the claim of the opening post: "The human elo of Stockfish with 20 kn/s is ~3000 according to this test"

20K nps is a litle light from my testing to get 3000 - I would say it's closer to 100k/nps give or take 40k/nps

[mehmet1921]

I made another test. Now time control is 10 minutes/game

Stockfish (10 kn /s) vs Junior 6 (5000 kn/s)

Program Elo + - Games Score Av.Op. Draws

1 Stockfish 271119 64 BMI2 : 2916 61 59 150 77.3 % 2702 14.7 %
2 Junior 6.0 : 2702 59 61 150 22.7 % 2916 14.7 %

(Fritz 11 Gui, 10 minutes/game, Core i7 9750h, Book2014 opening book, ponder off)

Individual statistics:

1 Stockfish 271119 64 BMI2 : 2916 150 (+105,= 22,- 23), 77.3 %

Junior 6.0 : 150 (+105,= 22,- 23), 77.3 %

2 Junior 6.0 : 2702 150 (+ 23,= 22,-105), 22.7 %

Stockfish 271119 64 BMI2 : 150 (+ 23,= 22,-105), 22.7 %

http://www.mediafire.com/file/hcd3kv4ki ... 5.pgn/file



The performance of Deep Junior 6 is 2702 at Dortmund 2000 tournament. (Hardware: 8 x Pentium 700 Mh)

https://theweekinchess.com/html/twic297.html#2

In my test Junior 6 % 60 faster than Deep Junior 6 (Dortmund 2000)


The human elo of Stockfish with 10 kn/s is ~2950 according to this test