To clarify, I use these book moves only for automated testing, to compare one version of my program against another. Not for any other purpose. The essential thing is that there is a variety in the opening positions, a large set (since I need to run a lot of games and don't want to repeat positions) and that they are not so unbalanced that one side is just going to always win.
--Jon
25,000 opening positions
Moderators: hgm, Rebel, chrisw
-
jdart
- Posts: 4366
- Joined: Fri Mar 10, 2006 5:23 am
- Location: http://www.arasanchess.org
-
Frank Quisinsky
- Posts: 6808
- Joined: Wed Nov 18, 2009 7:16 pm
- Location: Gutweiler, Germany
- Full name: Frank Quisinsky
Re: 25,000 opening positions
Hi Jon,
balanced or unbalanced is the big question!
After Stockfish and Komodo analyzed my database and I reject with 0.50 / -0.35 ... it seems for Houdini are more of 3% not balanced enough with 0.65 / -0.40. To create opening books with one engine analysis only can't be right.
If I thinking on evals from ICE / Texel / Fizbo / SmarThink after opening books moves ... it will be not lesser as 3%.
Eval from Arasan is very solid in comparing with other engines! Never I had Problems in optimization with Arasan games (means in the time I am working on my FCP Live Book).
With other words:
All the works we can do in find out balanced positions cann't be perfect. Often Fizbo and ICE are more right in eval as not right if I am looking after opening book in game progress. Texel is here very interesting too. SmarThink can find here very interesting defensive moves all the other TOP engines never can find. But unfortunately, SmarThink have tactical problems in the beginning of the game. To reject the particular moves from SmarThink games ... I am not strong enough for it.
In my interest is to create opening book with "balanced positions". So I am searching for all of the ECO codes ... 3 moves after ECO code was formed.
In reality ...
To a lot of ECO Code the result will be bad because it's not easy to find out balanced opening lines to: A51, A59, A64, A74, B85, C38, C58, C74, D62, E80, E85, E87 ... to other "unpopular" systems too.
In BookUp software better Chess Opening Wizard (COW) can be found very nice ideas ...
Maybe it make sense to looking with the final result I have in 6 months after 10 engine analysis with the professional version of COW!
Ferdinand gave the tip.
I am looking a bit and contact the author of COW.
Best
Frank
PS: Looking these days in Arasan with test-positions I collect all the years. Seems to be clearly improved since the last version I am tested for my "now" older FCP Rating List. Nice to see that again and again all the years you are working on Arasan. Often I wish me I had such an energy for one topic only. I don't have it ...
But what you do with Arasan must be good. Can be see very easy with each new version of Arasan.
balanced or unbalanced is the big question!
After Stockfish and Komodo analyzed my database and I reject with 0.50 / -0.35 ... it seems for Houdini are more of 3% not balanced enough with 0.65 / -0.40. To create opening books with one engine analysis only can't be right.
If I thinking on evals from ICE / Texel / Fizbo / SmarThink after opening books moves ... it will be not lesser as 3%.
Eval from Arasan is very solid in comparing with other engines! Never I had Problems in optimization with Arasan games (means in the time I am working on my FCP Live Book).
With other words:
All the works we can do in find out balanced positions cann't be perfect. Often Fizbo and ICE are more right in eval as not right if I am looking after opening book in game progress. Texel is here very interesting too. SmarThink can find here very interesting defensive moves all the other TOP engines never can find. But unfortunately, SmarThink have tactical problems in the beginning of the game. To reject the particular moves from SmarThink games ... I am not strong enough for it.
In my interest is to create opening book with "balanced positions". So I am searching for all of the ECO codes ... 3 moves after ECO code was formed.
In reality ...
To a lot of ECO Code the result will be bad because it's not easy to find out balanced opening lines to: A51, A59, A64, A74, B85, C38, C58, C74, D62, E80, E85, E87 ... to other "unpopular" systems too.
In BookUp software better Chess Opening Wizard (COW) can be found very nice ideas ...
Maybe it make sense to looking with the final result I have in 6 months after 10 engine analysis with the professional version of COW!
Ferdinand gave the tip.
I am looking a bit and contact the author of COW.
Best
Frank
PS: Looking these days in Arasan with test-positions I collect all the years. Seems to be clearly improved since the last version I am tested for my "now" older FCP Rating List. Nice to see that again and again all the years you are working on Arasan. Often I wish me I had such an energy for one topic only. I don't have it ...
But what you do with Arasan must be good. Can be see very easy with each new version of Arasan.