For Eternity I, I seem to recall that the winner took 18 months to get a solution. The puzzle has only been out for ~4 months.

“To the proper military authority: I’m calling to the commandant’s attention the recent changes noticed in the attitude towards fighting of soldier Daniel. He has attempted on several occasions to escape his duties … and made defeatist comments in public.”

:)

-Brian

Find the Eternity`s solution!!!

FREE DOWNLOAD
http://rapidshare.com/files/58807680/setupInfinity2-Demo.exe

WEB PAGE
http://www.infinity2game.blogspot.com

I’m just curious, how large is your search space at the moment (with your current heuristics)? Have you calculated anything on computing time / position?
Whats your general approach? Which algorithms have you discarded? Have you tried any genetic algorithms? I heard saw someone mention simulated annealing as something interesting to try.

I’m working on a solving engine written in c++ using stl/boost and a java gui on the eclipse platform which connects to the engine via jni.

My goal is to have a platform to try out different heuristics and compare them against each other at various board sizes to get a feel of their strengths and weaknesses. I’m setting up a testbench for the engine which can generate random boards of any size – both solvable and none solvable, so I can verify my code.

I do not think I will take home the money though, the eternity2 project at eternity2.net has probably already a solution to this, or will have in the coming months – at least before me.

Regards,

Daniel

Well, it’s been quite a while since the puzzle was published. I haven’t dug around too much but I imagine someone has sent in a valid solution by now. But if not, or if you still want to hear my weird ideas for attacking it, drop me an email.

Hope all is well in beantown,
Robby

Good luck on the puzzle, Benji. Keep us posted on how your endeavors are going.

-Brian

I am kind of on the same page as you regarding the reward system of eternityii.net’s distributed computing effort. I programmed a solver a few weeks ago (before the game came out) that does about 20M piece placements per second with lookahead (checks whether there is any empty board cell with zero playable pieces to avoid wasting time). I was considering setting up a distributed computing project where rather than two people splitting the prize, it would actually be distributed quite a bit more. Something like this:

1) $400K goes to the developer of the distributed application
2) $400K goes to the person that actually finds the solution, if applicable :-)
3) $400K gets split among the people that complete the most work units (i.e. the more work you do, the more you get)
4) $400K goes to a charity selected by the folks that do at least one or two work units.
5) $400K gets split at random among people (say, 400) that do at least one or two work units (e.g. like a lottery)

OK, this payout scheme certainly way more complicated than a simple 50/50 split, but it provides an incentive to a lot of people to participate that wouldn’t otherwise. #3 provides an incentive to folks with lots of computing resources to work the problem even though their chances of finding a solution themselves is still quite small. #4 would be great for generating publicity and rallying folks who’d rather see their computer cycles going toward helping a worthy cause. #5 gets people with less computing resources involved because they know that they still have a chance at getting $1000 or whatever just for participating minimally.

Of course the amounts could be shuffled around and a tax attorney would have to weigh in to ensure that things aren’t taxed a zillion times as the money is divided, but in principle, it seems like a more “fun” way to work the problem. I think it would just be cool to watch the online graphs change over time, showing how many work units different people have done and stuff like that :-) I was looking at just using BOINC, but hadn’t really gotten that far.

The issue with requiring folks to enter in their own piece definitions makes things a bit more burdensome on wouldbe participants. CM’s threats aside, I don’t really know what the copyright laws would prevent as far as including an encoding of the game pieces in the software (to what extent would that be a derived work, especially if the internal encoding is nothing like 256 pieces with four edges). But regardless it seems reasonable that the judges could disqualify you for whatever reason they want to, so even if no laws are broken, you lose the prize if they somehow “found out”. Makes it all less conducive to a public distributed computing effort.

Anyway, you probably know now that Toy R Us in the US has it for quite a bit less than $120, but maybe those few extra days let you find your solution? :-)

Cheers,
Bill

What would you say are the odds of someone solving this puzzle manaully without any computer model?

Monde