| « home | Browsing all comments (spoilers: on | off) (info: on | off) | next 25 |
 |
<Stephan> (SPOILER) This level is rather loose; I suspect there are several possible ways to arrive at the finish. |
 |
<Doom> (SPOILER) I feel a little stupid because it's possible to push the last block left instead of right reducing the solution style. Anyway, that's not worth making another version IMO. |
 |
<Doom> (SPOILER) Added speedrun: 60 moves (old: 74). Faster version of the original solution <Stephan> (SPOILER) Nice one nLogic! I knew it might be possible to solve it that way, but I hadn't done it yet. |
 |
<Tom 7> (SPOILER) After realizing this problem is substantially different than Ricochet Robot, it was pretty easy to solve with a special-purpose solver. I sank several hours into trying to solve this one by hand, though. Definitely very, very hard. I want to experiment a bit of my own with autogenerated levels. |
 |
<nLogic> (SPOILER) Cook: Didn't use the yellow bricks. |
|
|
<nLogic> (SPOILER) Added solution "Alternate": Completely different from the original solution, but exactly the same number of moves. <Stephan> (SPOILER) This is a simple level if you know the secret. It looks really simple but is actually full of skulduggery and deception. |
 |
<Stephan> Very impressive solution, iain! |
|
|
<Doom> It should be impossible to do it without moving the gold blocks around first. I also replaced hugbots with daleks just for fun. |
|
|
<Tom 7> (SPOILER) Cook: Ditto! <Doom> (SPOILER) Whoa, that's a clever cook... I guess I have no choice but to make another version. <Stephan> (SPOILER) Cook: was able to solve without using gold blocks. |
 |
<OCTweak> More than 3 weeks trying this one and the close I can get is one move to the solution. I'm not giving up but knowing there is another one thougher than this one (Great Waste of Time) makes me crazy! 8-) |
 |
<Tom 7> (SPOILER) Thanks -- like I always say, the best thing about torture is that it feels so good when it's over. A good way to check if your solution is actually a cook is to download the original solution and compare. You can do this by hitting ctrl-m on the load screen after you've solved a level, and then pressing ctrl-d to download all of the solutions. Also, if you have a solution that's 16 moves faster than the speedrun, you should upload that new record! <pieguy> (SPOILER) okay, sorry for ruining the suprise. when my first-try solution was 16 moves shorter than the speed record, i just sort of assumed i had cooked it. excellent level, then. it's the kind of level i would create to torture people. <Tom 7> (SPOILER) Don't read the spoilers! ;) Well, I am glad you like it, at least. =) <mjn> (SPOILER) !!!!!!! Oh my god, that's awesome. I've been tearing my hair out over this one for months; I never once thought of that... I think this definitely wins the most-evil prize. |
|
|
<Tom 7> (SPOILER) So does it make levels starting from the solved state and 'retracting' to find the most distant precursor? I made some stuff for the escape level editor that does retractions for gold blocks and spheres (I don't think anyone uses it), but never thought to do *search*. Good idea. |
|
|
<Tom 7> (SPOILER) Nope, that's the only solution as far as I know. <pieguy> (SPOILER) is this level cooked? you can flip the exit, then push a block on top, and flip it over again. i haven't tried to solve it the "obvious" way. |
|
|
<pieguy> (SPOILER) it turns out that for the program i used to create this level, finding the best solution is no more than a side effect of creating the level itself, with one minor difference. my program assumes that when you push a block, you move into the square previously occupied by the block. this allows me to only consider levels with even parity. for this particular level, 15 minutes of computer time was required (on my 2800+ athlon), and about 300 MB of ram. my program places the blocks in such a way to ensure the longest possible shortest solution. <Tom 7> (SPOILER) Thanks, mjn. Yeah, my solver has hashing, but I had it turned off because it made things a lot slower for Ricochet Robot solving. But I guess I was expecting a much shallower solution! I'll try again.. <mjn> (SPOILER) Whoops, I'm an idiot--of course the solver's only going to find one solution if duplicate positions are ignored; it won't be able to duplicate the ending position! Never mind, I have no idea what the minimum complexity is. The minimum number of block pushes, though, should be 68. <mjn> (SPOILER) Heh heh, I gave up and wrote a solver as well. My solution has a complexity of something like 27, including the first move. (I got that number by watching the solution replay; I might be off by one or two.) Also, unless there's a bug in my program, that's the *only* solution (except for a couple minor variations towards the end). Does your solver check for duplicate positions? Mine was threatening to take a long long time, too, but fortunately, adding a hashtable to check for repeats cut it down to a few seconds. |
| « home | Browsing all comments (spoilers: on | off) (info: on | off) | next 25 |