Solution: Unsafe

Answer: MATRIX

Written by Brian Chen and Jakob Weisblat

This is a puzzle about the 2012 train safety PSA, "Dumb Ways to Die".

This is clued obliquely a few places in the puzzle:

• The description of the player starts with "You look like you could be in a cartoon."
• The map in the train station has the Melbourne Metro insignia on it.
• There is an old screen in the Control Room that is playing a PSA about train safety.
• The song that plays through the earbuds is a vague description of the song, mentioning that it was a campaign and went viral a few years ago.
• The player is called "dumb" in various places - upon reaching any of the deaths labelled "dumbest" in the video, and upon trying to poke the bear without a stick, among elsewhere.

Solvers will proceed through the puzzle, eventually reaching one of the 14 endings. Upon reaching an ending, the game tells you the number of moves, as well as the optimal number of moves. This indicates that the solver should be looking for different endings, as well as the fastest way to each ending.

Here is a table of each "dumb way to die" from the song, whether it is possible to reach it in the game and one possible optimal sequence of moves.

The death text implied heavily that the length of the optimal path to a death is relevant. Making a table of the lyrics corresponding to deaths, in order of the song, solvers will notice that the length of the path is simply the number of letters in the lyric. So we won't just index the path length into the song lyric. However, solvers may notice that every single optimal path includes the suspiciously named Tunnel of Extraction exactly once. Indexing into the lyrics by the position of the move that leads to passage through the tunnel (or, for one ending, the crash-landing in the tunnel) gives:

This yields the cluephrase ARRAY OF NUMBERS, which is a MATRIX, the answer to this puzzle.

Author’s Notes

Brian: This was one of my more ambitious puzzle ideas. I was originally planning to save it for Mystery Hunt or later due to the scope, and due to the source material being used in this year's Mystery Hunt. However, in early February, the GPH editors decided that we wanted more ambitious puzzles and my other puzzle ideas weren't panning out, so I decided to write it. I know of a fair number of text adventure puzzles that had already appeared in puzzlehunts (e.g. MUMS's Adventure, Caltech's Hidden Verbiage, or of course any of the text adventures that have appeared in the MIT Mystery Hunt), but wanted to make a puzzle out of a slightly more open-world and traditional pick-up-everything-and-figure-out-how-to-use-it text adventure. I also felt like Dumb Ways to Die could be further explored after the Mystery Hunt puzzle. Finally, I thought it would be funny to include a blatant Tunnel of Extraction in the universe and try to weakly spin it off as something that made sense in-universe.

The somewhat involved extraction method was chosen to make sure that we could tell teams when they had reached an optimal sequence of moves, while teams could not short-circuit too much of the puzzle by reverse-engineering the puzzle to find those optimal numbers (which is difficult, but far from impossible, and we know of at least one team that immediately reverse-engineered the puzzle when they opened it). This did require a fair amount of sacrifice in terms of how much finagling of some of the endings and padding them out with the exact number of moves was necessary (most egregiously, of course, the mechanism by which you have to press the button 18 times to call a train, which was initially a stopgap but was kept in because it didn't feel worthwhile to pad three endings with identical complex sequences of 18 moves). It also required a lot of caution around the topology of the Tunnel of Extraction. If we had more time I would have loved to come up with more natural ways to force the optimal move sequences. We tried to make the puzzle primarily about finding all the different endings and not about optimizing getting there, so we tried to avoid having not-quite-optimal paths.

Jakob spent a good week of his life implementing endings for this puzzle. In the end the source code was around 15.5k words.

Another part of making the move sequences exact was that we generally had to try to make Inform 7, the text adventure engine, less smart – e.g., we had to disallow trying to go through closed doors (which would by default try to open them first in the same turn) and taking multiple things with "take all", so that move count was exact and we didn't disadvantage teams who got almost perfect runs but might not realize these optimizations. (There were a few Inform shortcuts we were ok with keeping – e.g., walking off stools.)

Jakob: Going into the hunt, we believed there was a 70% chance that all of our endings were optimal, but several problems were discovered by teams that we had to fix during the hunt:

1. We had forgotten to require that you are holding a key in order to unlock a door with it (Inform is willing to let the player pick up a key in order to unlock a door with it on the same turn).
2. We had forgotten to include in the definition of "dip" that you can't dip a bucket in a closed vat.
3. We misspelled "unrefrigerated" (as "unrefridgerated"), resulting in the optimal number of turns for that path being one more than the number of letters. We had to fix this during the hunt and made the inner safe automatically close after you put something in it. However, it didn't affect the index and many teams seemed to have solved it without noticing.
4. Finally, we didn't notice that players had some flexibility in taking off their belt and hat before or after passing through the tunnel in the piranhas ending without sacrificing move optimality. This made the index for that route ambiguous, but since we only learned about this on Sunday, we assume that not a lot of teams were affected.