Fabien, good question(s).
Its based upon a fuzzy idea (also triggered by one of your comments, than maybe Draughts is about who promotes first, and than go from there...).
Yes I was wondering whether draughts strategy is mostly about making a king first. With the implied effect that you cannot lose if you do that without giving away too much material. But even assuming that's true, it still leaves open the question of whether the opponent can get a draw. I didn't give that much thought, but I can imagine some threshold, say 7 plies, that would go like this:
IF I make a king first
AND my opponent cannot do the same <n> plies later
THEN I win (more often than not)
So I was thinking could we (one way or the other), combine BT DBs with search results in normal 10x10 Draughts.
From a search point of view , 2 different moves could look equal, but if one yields (according to the DB) a faster promotion to king, it could be worth trying.
As BT DBs are easier to generate, we could generate these beyond the piece limits (=8) we have today.
Maybe to 10 (so 5x5), as you suggested.
My plan was related. I wanted to learn a BT eval, and see if it played decent draughts during the middle game. I did learn an eval and it was distinctly better than a generic one, but I can't plug it in just like that, because it has no king evaluation (even material). Then other subjects were opened, and I switched to working on the protocol and cleaning up my GUI code. I should run a simple experiment anyway with a fixed value for kings, as it's not much work and the result can affect your project. I expect the BT eval to do badly: give too much to make a king first.
BT could also help me putting Killer draughts into perspective. I see an imaginary "drawishness" axis like this:
- normal draughts: too many draws
- killer draughts: good compromise?
- BT: no draws, but is it still draughts?
So if BT "fails" to be draughts-like (although it could still be fun), this would reinforce Killer draughts as a good replacement candidate for the official rules.