Gabor Szots wrote: ↑Mon Apr 01, 2019 12:34 pm
pedrox wrote: ↑Mon Apr 01, 2019 12:05 pm
At the moment I will stop working on Caligula, if there is a new version it will be 1.0
The next version will be from DanaSah.
It's good to hear you haven't abandoned Danasah, which I was afraid of.
What's the main difference between the engines, besides Caligula being WB only?
DanaSah is more complete than Caligula.
DanaSah has versions at least for windows, linux, android, pockectpc and mac. It has protocol xboard and uci. It can play endgames with the Scorpio bitbases and Gaviota tablebases. It has a version that play with less strengh. It can play Chess960. Caligula only has version now for windows, only protocol xboard, only uses the bitbases of Scorpio and does not play FRC.
DanaSah uses in the programming a board that we could say is 8x8, while Caligula uses a 12x10 board to facilitate knowing when a piece in its movement leaves the board. This board is used in the generation of moves, in the function to know if a piece is attacked, SEE, mobility, etc. Thanks to this the code is smaller and more compact. But the interesting thing was to know if it could be faster and in my case I can say that it is not faster. This was the main reason to do Caligula.
In the search function there are changes, Caligula is generally more aggressive with the prunning. Surely something superior.
In the evaluation function, it uses different values for psqt, movility, tropism, etc. There is more weighting between the midle game and endgame. Possibly Caligula's evaluation is also better.