Position evaluation database

[Bozon]

I was thinking about creating a database containing engine evaluations of positions. I would like to include evaluations of several engines (at least the 10 top engines) and I was thinking about some web interface where the user can select the engines he wants to consider and can also submit his evaluations (with a documentation of the parameters used; the engine name, version number and parameters could be combined into some unique hash). Does there exist something like this already? If not, I would be interested in all ideas/comments you have. I'm not an expert so I can't tell but if there is some interest in this, it would be great to team up and realize this idea.

One problem I already want to address is that evaluations of different engines are not comparable a priori (think of the Stockfish evaluations which are usually higher than those of other engines). I therefore though about analyzing "move qualities", i.e., I consider a position which has score x and then I consider a move and the resulting score y and take x/y as the move quality. The move qualities obtained from different engines are then comparable. There are many more problems to deal with of course.

[Guenther]

I was thinking about creating a database containing engine evaluations of positions. I would like to include evaluations of several engines (at least the 10 top engines) and I was thinking about some web interface where the user can select the engines he wants to consider and can also submit his evaluations (with a documentation of the parameters used; the engine name, version number and parameters could be combined into some unique hash). Does there exist something like this already? If not, I would be interested in all ideas/comments you have. I'm not an expert so I can't tell but if there is some interest in this, it would be great to team up and realize this idea.

One problem I already want to address is that evaluations of different engines are not comparable a priori (think of the Stockfish evaluations which are usually higher than those of other engines). I therefore though about analyzing "move qualities", i.e., I consider a position which has score x and then I consider a move and the resulting score y and take x/y as the move quality. The move qualities obtained from different engines are then comparable. There are many more problems to deal with of course.

There was a huge project called CAP(or C.A.P. = Chess analysis project) which lasted for years and some software later even imported the data into opening books e.g.
Chess Assistant. IIRC Crafty also could make use of it. I don't remember too much details now as it all was quite long ago.
The archives say the start already was announced in 1999 by Dann Corbit, who was a very active member here until 2011.
http://www.stmintz.com/ccc/index.php?id=45105
A search in the forum and googling about CAP should find a lot.

Guenther

[jdart]

ChessBase also has an on-line db of engine evals now (need a Playchess account to access). A single position can have multiple associated evals from different engines.

--Jon