Killer draughts

[Fabien Letouzey]

Hi all,

This theme appeared in a couple of threads but I thought it might deserve its own.

First the links to recent discussions:
http://fmjd.org/bb3/viewtopic.php?p=114524#p114524
http://fmjd.org/bb3/viewtopic.php?p=114607#p114607
http://fmjd.org/bb3/viewtopic.php?p=114652#p114652

---

I ran a little experiment about the draw rate: Scan against itself.

normal draughts: 87/95/99%
killer draughts: 39/50/69%

The time control is 1/10/100s per game (+ 1% increment), single core. Games were adjudicated by endgame tables, which only included the 3 vs. 3 subset of 6 pieces (and all smaller sets). I don't have enough memory for the full set on 4 simultaneous experiments, and I don't have code for disk-based tables.

Tournament conditions would correspond to a few thousand seconds, so we can only extrapolate from these data. A more realistic experiment would use DXP games and different engines of similar strength, which are not yet publicly available for killer draughts.

In any case I reiterate: although I am in favour of switching to killer draughts, it's not going to solve the draw problem. It's just going to make draughts more in line with other games with draws. In top-level chess for instance, draws are already becoming a problem: http://www.talkchess.com/forum/viewtopi ... _view=flat

Games with no draws (e.g. the first one to promote wins as in breakthrough: https://en.m.wikipedia.org/wiki/Breakth ... oard_game)) seem the only long-term solution. Even then, I guess approaching perfect play would bring its own problems for the losing side.

Fabien.

[BertTuyt]

Games with no draws (e.g. the first one to promote wins as in breakthrough)

Facinating thought!
At least EndGame DBs would become very simple, only win/lose, and only databases with man are required.
Did not do the math, but it would be interesting to check what the maximum DB would be, with current technology.

Maybe perfect play would become possible within the horizon, and than also this option does not makes sense (at least not for computers).
Any thoughts what the outcome would be, white starts and white wins?

Bert

[Rein Halbersma]

Games with no draws (e.g. the first one to promote wins as in breakthrough)

Facinating thought!
At least EndGame DBs would become very simple, only win/lose, and only databases with man are required.
Did not do the math, but it would be interesting to check what the maximum DB would be, with current technology.

Maybe perfect play would become possible within the horizon, and than also this option does not makes sense (at least not for computers).
Any thoughts what the outcome would be, white starts and white wins?

Bert

Martin Fierz (of Cake & Checkerboard fame) has solved the break-through variant for 8x8 checkers. He called it Kingscourt: http://checker-board.blogspot.nl/2006/0 ... olved.html

[Fabien Letouzey]

Facinating thought!
At least EndGame DBs would become very simple, only win/lose, and only databases with man are required.
Did not do the math, but it would be interesting to check what the maximum DB would be, with current technology.

Maybe perfect play would become possible within the horizon, and than also this option does not makes sense (at least not for computers).
Any thoughts what the outcome would be, white starts and white wins?

Bert

I didn't mean to suggest this as a serious proposition, and was more thinking of games like Shogi and Go where draws are avoided by the rules. In draughts, one king vs. 3-4 men is an interesting imbalance and makes the game richer IMO. Regarding endgame tables, I had the same thoughts as you, and guessed that adding 2 pieces would be possible.

However I think that removing kings would change the game so much that current engine design wouldn't be suitable anymore. For one thing, one can use technology similar to breakthrough tables in Dragon to find wins much earlier than with (full-board) endgame tables, thanks to locality and the limited possibilities of counter-play (race only). And maybe PN-search/MC methods becomes better on this type of game (no draws, possible early wins). If you want to know more, look at breakthrough (the game) publications, E.g. "Solving breakthrough with Race Patterns and Job-Level Proof Number Search".

[BertTuyt]

I didn't mean to suggest this as a serious proposition

I tend to agree, but at least it is an interesting challenge....

Bert