brainwagon

Martin Fierz, author of the truly excellent checkers program Cake, has released a checkers program in Java. I run Cake on my PC, and also at times under Wine on Linux, but it is nice to have a version which can run on Mac/Linux without any hassle.

jCheckers.

For fun, I fired up Milhouse and played a game against jCheckers. I gave each engine five seconds to think (which is pretty low, but it makes the game go fast). Milhouse managed to eek out a draw, but it appears to me that it was lucky: Milhouse appeared to be in a pretty cramped position for much of the game. It seems to me that Milhouse narrowly broke through the log jam at the end, and each side ends up with 4 Kings, which is a drawn position.

[Black "Milhouse"]
[White " jCheckers"]
[Event "First Game Against jCheckers"]
[Result "1/2-1/2"]
[Date "2010-03-28"]
1. 9-14 22-18 2. 10-15 18x9 3. 5x14 25-22 4. 15-19 24x15 5. 11x25 29x22 6. 8-11
28-24 7. 6-10 22-18 8. 1-5 18x9 9. 5x14 26-22 10. 11-15 31-26 11. 3-8 23-18
12. 14x23 26x19 13. 7-11 21-17 14. 11-16 27-23 15. 8-11 17-13 16. 16-20 32-27
17. 11-16 13-9 18. 4-8 30-25 19. 8-11 9-5 20. 2-6 5-1 21. 6-9 25-21 22. 9-14
1-6 23. 14-18 23x7 24. 16x32 6-10 25. 20x27 10x19 26. 27-31 19-23 27. 31-27
23-19 28. 11-16 19-15 29. 16-20 15-19 30. 32-28 22-18 31. 20-24 7-3 32. 27-32
3-8 33. 24-27 8-11 34. 27-31 21-17 
1/2-1/2

Here’s the graphics from jCheckers, showing the final position.

Final Position

Addendum: Here’s one of the interesting position from the game. Cake finds a move which wins a checker, but both jCheckers and Milhouse seem to struggle to search deep enough to find it. Jcheckers played 14. … 27-23 which appears to be a draw, but 14. … 17-13 is a better move, and should win at least a checker.

White to play and win (jCheckers missed it, and so does Milhouse)