Shadowcat on 25/6/2010 at 14:08
Further to the Malevolence Engine "losing patience" with Robot Police, I just ran it over all of my machines*, and I notice that (*** Judiciary SPOILER... ***) it's perfectly happy with Judiciary, which is interesting, because my approach to Judiciary was to copy my Police machine, cram it into the top half of the grid, and use its output as my starting point for the new problem. (I rather suspect that most people did much the same thing.) So I really do think its complaints about Police are probably bogus (unless there's an O(n) solution that I'm just not seeing, but damned if I can think how that would work...)
(*) I had one failure -- I hadn't handled empty strings for Politicians -- and it also "lost patience" with Robomecha, which I quickly realised was entirely fair, given that my algorithm was O(n^2), and I could see that O(n) was perfectly do-able with a much smaller machine. Quite the blunder, that one :)
Nameless Voice on 25/6/2010 at 15:11
I noticed that parallel between Police and Judiciary too, for the same reason, and came to the conclusion that each test must have had an acceptable time set up, and that either the one for Police is too low, or I'm missing something.
hooded_paladin on 25/6/2010 at 23:00
Shadowcat - I was wondering if everyone else did that for Judiciaries, too. I'm not too good at packing so I had only narrow sections left and I was having a hell of a time placing the last parts to that problem. Blegh. And then Seraphim was a joke.
Al_B on 25/6/2010 at 23:46
It's what I did too. Since the problem involves dividing a pattern in half it seemed the sensible approach. You're right about Seraphim, but Ophanim is giving me significant headaches now.
Shadowcat on 27/6/2010 at 03:39
Quote Posted by Al_B
The other half [of the fun] is trying to match the results of other people here
Indeed! Even though I'm not planning to attempt to minimise pieces on all levels, several of mine definitely needed some attention. My original Rocket Planes was a bit of a beast at around 54 pieces. I was able to reduce that to 37 with some sensible re-routing, but clearly my approach was fundamentally inefficient when other people had 21 piece and 19 piece solutions, so I started over. That led me to a nice 21 piece solution, and I managed to cut out another piece to take it down to 20, and then scratched my head for a bit before figuring out how a few changes enabled a 19 piece version.
So yeah... thanks, you guys :) I can't reduce the numbers already posted for any of the puzzles, but it's been fun trying to match them!
Al_B on 28/6/2010 at 06:50
I sent the link to the game to a friend and he came up with a 16 piece rocket plane solution...
Shadowcat on 28/6/2010 at 08:06
Damn him.
dvrabel on 28/6/2010 at 09:32
What algorithm have people used to solve Rocket Plane? I used a bubble sort and a machine with four states -- INIT, B, R, R_SWAPPED (B and R indicate the color to write out next. This requires 15 active pieces (writers and branches) and at most 13 conveyors.
