What previous tries already indicated there is ELO in big data and perhaps more I thought at first sight. In the first YAT edition ProDeo 2.0 scored an 78 ELO improvement with as only change a computer generated extended opening book made from CCRL and CEGT games of about 9 million positions, about 100 Mb uncompressed, see 2016.cht in the BOOKS directory.
While 78 ELO is probably a bit overrated (and the next edition of YAT will tell more) it further awakened my curiosity to invest more time and energy in computer generated ELO and as such I started the *.EBF project. ProDeo 2.1 now supports the new *.EBF format and part of the analysis is included and will be further updated.
Click to enlarge
If you are the owner of ChessPartner then you can watch how the various opening book systems smoothly work together via the engine -> tutor window.
A short 40/60 1000 game match ended in favor of the new EBF format with 53.7% indicating an 25 ELO increase from the 2275 produced book hits. More below.
Moves played from Programmable Book : 0
Moves played from Tournament Book : 0
Moves played from Main Book : 0
Moves played from EBF Book : 2275
Moves played from CHT Book : 0
That if there is so much ELO in this unexplored area I prefer ProDeo 2.1 (and on) to be tested as released. It's the same 2.0 engine anyway. As such I have made some changes, the standard opening book is hardwired, book learning also, you can't turn it off, the permanent brain command by the GUI is ignored as well, for serious testing I want to play the thing as its best possible and not crippled removing opening books and learning facilities on rating lists.
And as such I want to give a signal too:
There is ELO in big data.
Programmers immediately can profit from the large 2016.cht (extended) opening book in the BOOKS folder by implementing the *.CHT source code as introduced in Prodeo 1.84.
Programmers can profit from the *.EBF project by implementing the *.EBF source code and download the upcoming *.EBF extended opening books.
Download the the newest available 2016.ebf and extract it into BOOKS folder. All positions are analysed with Stockfish 7 at 2 seconds per move. The PLY=16 version is included in the ProDeo 2.1 download.
For the meaning of the CHT parameters we refer to Prodeo 1.84.
The [EBF Priority = ] parameter decides which book should be used as first one. For the moment the CHT book performs better and thus the CHT setting is used. When a move is found in the CHT book it is played, if nothing is found it will consult the EBF book. Using the setting EBF will reverse the search order.
Follow us at Facebook for the latest developments.