Page 1 of 1

Project idea

PostPosted: 03 Feb 2006, 10:20
by Joerg Nowak
Hello chess960 engine programmers,

I plan to initiate a project with distributed computing and Chess960.

Goal of the project:
- create a non commercial web database with millions of Chess960 games and all positions (may be estimation of each position)
- develop huge opening book(s) for all start positions

Project Framework:
BOINC
BOINC is a project framework used by the SETI (Search for extra terrestic intelligence) project. BOINC supports now also other non commercial projects.
This framework supports distributed computing on many (millions) of clients

First steps:
- forme a team to start this project (especially programmer(s) of strong chess960 engines, C++ skills, knowledge on different patforms (Windows, Linux, Mac), database know how, chess know how).

- Find sponsors

Who is interested to join such kind of project ?
In the project we need a programmer who is able to adapt his own source code of a strong engine or a free source code of a strong engine to the BOINC framework.

You can read more about this framework of distributed computing on
http://boinc.berkeley.edu/



Please contact me on www.chess-960.org (info@chess-960.org)

Re: Project idea

PostPosted: 10 Feb 2006, 08:00
by jshriver
Last year I did something similiar but not in a distributed environment.

Basically I just took the full FRC opening variation list. And let 2 engines (think it was gothmog and forget the other) and let them play 20 or so games saving the games in pgn.

Then load another opening position from a FEN string.

Problem is... limited FRC engines, and not sure if any are as strong as Fruit, Crafty, Rybka for giving great positional plays.

If you did just 1 opening per day it would take years to create even a 20 game per opening book.

Maybe if a distributed attempt was made it would be very useful.
If you had 960 participants, then if each ran for a week you might get a nice size book.

Sounds like a fun project, I hope you do well. If you get a client up and running let me know I'll gladly run it on a couple machines.

-Josh

Re: Project idea

PostPosted: 10 Feb 2006, 08:02
by jshriver
Pretty site, but do you have an english translated version? Would love to read it.

Josh

Re: Project idea

PostPosted: 10 Feb 2006, 08:22
by Mihail Croitor
Pretty site, but do you have an english translated version? Would love to read it.

press GB flag under menu.

Re: Project idea

PostPosted: 06 Apr 2006, 10:56
by Joerg Nowak
I'll give a short update regarding the project chess960athome:

Now we have formed a small team to realise the project.

First we installed and compiled the BOINC server platform on the project server (http://www.chess960athome.org/alpha).

We found 200 alpha testers (it was much easier than expected).

Now we are in the phase of developing the BOINC client application.

Our idea is to calculate one move in a work unit (WU) by a adapted free (UCI) chess960 engine with a fixed depth (first try).
Then we want to use this move to create the next work unit (BOINC terminus).
I hope we are able to realize that in the next month.

But at the end (if everything works) we 'll have a string like:

position fen qbbnnrkr/pppppppp/8/8/8/8/PPPPPPPP/QBBNNRKR w KQkq - 0 1 moves d2d4 d7d5 e1d3 [... many moves]

One of our problems is to convert such a string to the pgn notation format what we need to publish the game and put it in the database:


1.d4 d5 2.Nd3 ...


Has somebody such a piece of code to convert to pgn format ?

Re: Project idea

PostPosted: 09 Apr 2006, 21:16
by jshriver
I'd be interesting in running your code, so count me as alpha tester 201 :)

-Josh

Re: Project idea

PostPosted: 09 Apr 2006, 21:30
by Joerg Nowak
The client application isn't ready yet.

Re: Project idea

PostPosted: 10 Apr 2006, 10:39
by Pradu
Joerg Nowak wrote:Has somebody such a piece of code to convert to pgn format ?

If you are looking for changing coordinate notation chessmoves into SAN (the hardest part of generating a PGN IMHO), you can look at the source of a lot of open-source engines. My program can change moves into both SAN and chessbase (non-optimal) SAN but my program is in bitboards. Mine has castling moves stored differently internally than coordinate notation moves so you won't have a problem changing the code to support FRC castling. You might want to look for a different engine's source if you want a non-bitboard solution. Many winboard engines that support the WB2 setboard command (including mine) aslo have FEN parsing capablilties. But I suppose how you generate your PGN or parse FEN depends a lot on your internal board structure.

Re: Project idea

PostPosted: 10 Apr 2006, 13:21
by Joerg Nowak
For me it doesn't matter the code is based on a bitboard solution or not because I don't have a own engine.

I was hoping somebody has written such a tool before for private use.

What I search for is a batch program like "ucinotation2pgn" < [input_file] > [output_file]

input_file:
position fen qbbnnrkr/pppppppp/8/8/8/8/PPPPPPPP/QBBNNRKR w KQkq - 0 1 moves d2d4 d7d5 e1d3 ...

output_file:
[optional other header tags]
[FEN "qbbnnrkr/pppppppp/8/8/8/8/PPPPPPPP/QBBNNRKR w KQkq - 0 1 "]
1.d4 d5 2.Nd3 ...