Runaway's Revenge

[gfoot]

Despite (or because of?) it being one of the easiest puzzles on the site, Runaway Robot is also one of the most memorable - it was still challenging, but very accessible, and the levels scale well enough to break certain algorithms at staggered intervals and force you to work out better ones. The difficulty curve just makes it seem well-balanced.

I guess it's looked-down-on because there is a fairly clear "correct" solution, which by level 513 I'm sure everybody is using. I can see that if you thought your way to this solution right at the start, it would be a dull game, but for those of us who only got there one step at a time, I think it was more fun. I might be alone, but it's a puzzle I look back on fondly.

So it would be nice to resurrect it with a sequel, extending the rules enough to break the natural algorithm, forcing people to adapt in the same kind of way. I think the instruction limit and looping works well, and feels fundamental. The obvious thing for me is to borrow some ideas from RoboRally - e.g. conveyer belts and spinners - and maybe things like pushable blocks, which all remove your ability to condense the board.

I haven't thought about it enough to know whether this would actually make a good puzzle, or whether it would just be something that's really inefficient to solve with no real hope of finding an efficient algorithm. Adding rules like this is a bit contrived.

[adum]

hey gfoot. i've thought about the idea of making a version 2 of runaway robot too. i think it could be interesting, and there are a lot of modifications that could make it more challenging to solve. so far i've chosen to just make new puzzles instead of a version 2 of RR. but if someone wants to design it, i would be happy to put it up...

[gfoot]

I wonder if we could do something more interesting than just looping the input program, such as requiring the solution to be an HVM program which prints "R" or "D" until the robot crashes or escapes. This makes the looping behaviour less fixed, and allows the puzzle to be generated with a more complex solution. I'm not suggesting that the HVM program would be a complex solver - just a way of encoding more complex looping behaviours, making the puzzle harder to analyze.

However, it might again be just too hard to analyze efficiently this way.

[Nick-Aotmzgin]

i love puzzels, and bots can be active and helpful in the game itself.
it sound so brilliant.