Marc Lacrosse wrote:At the opposite, most UCI engines rely on the GUI for book operations (I know that Yace is a brilliant exception !) and what I miss mostly is some kind of book management by polyglot.
I read somewhere that you intended to add book-features to Polyglot.
Did you progress in this way ?
PS bookthinker is not a satisfying answer for book management; not only its use is a pain, but it cruelly lacks user-book features like the ones of crafty, yace or rebel (to name a few)
There is an experimental version that supports only NAIVE book making, I am afraid not as good as BookThinker. I had in mind to improve it before a release but then switched on to working on something else (that was in February).
It is now clear: I will release that version "as is" (but stable) next week, and try to improve things only afterwards.
Basically I don't want to spend time researching opening-book heuristics.
However, if one (yes that's YOU) tells me EXACTLY what he wants, I can do it hopefully in little time. Please open a thread in that case, thanks.
As an example, Leo mentionned a heuristics regarding minimum win % (separate for white and black, e.g. 50 and 45) for moves with many games (say at least 100 or 1000).
I have nothing against user books, just make sure I can figure out EXACT move probability from your data (you tell me how).
However I fail to see in which way PolyGlot would be easier to use than BookThinker.
Fabien.