Wow wow wow. 1h30 to crack it, really ?
I'm just done (100%) with probably 8h in Java, so I'm feeling bad now Were the rules and tests of the contest exactly the same as the single game we now have ?
Either my algo was good, and I was bad at implementing it, or my algo was bad.
I'm basically adding all "forced links" from the original state, and if the game is not solved after that, taking hypothesis as deep as needed to know if the hypothesis was good or bad (keeping in memory the state at each recursion).
Took me about 750 lines (lots of comments and quite verbose though).
No timeout and clean, but so far from 1h30 !
So, bad algo or bad implementation ?