SC2SEA.com - Starcraft 2 SEA eSports Community Site

This past week was the fifth birthday contest, a four-day puzzle hunt. The winner was Pox, who solved almost everything in the final 24-hours.

Worked solutions are now available at the puzzle pages here, and a full guess log here.

The following are some comments about the design and thoughts behind how the puzzles turned out.

Overall format
As several people have deduced, the format of the puzzle hunt is heavily inspired by the Melbourne Uni Maths and Stats Society - a series of puzzles released day culminating in a metapuzzle on the final day which assembles the answers of all the previous puzzles together to give the final solution. It has a lot of benefits in that you aren't blocked from progressing by being stuck on a single puzzle.

There are some differences - the MUMS puzzle hunt is supposed to be done in teams, there are prizes for fastest meta solve as well as two runners-up by points, where early solves (before hints) of regular puzzles are rewarded. It is also about a billion times more difficult.

I had intended a point system, but with a first-past-the-post meta puzzle solve with no runner-up prizes, I couldn't think of a way to do that without basically eliminating people who hadn't solved everything on release day from prize contention. A prize by point leaderboard meant if you fall behind, there would be no way to catch up by solving the meta more quickly.

The final result of this contest, Pox winning by solving all the regular puzzles on the last day or so, and then solving the meta before everyone else was a bit unfortunate in timing, but there was a tremendous amount of quid-pro-quo cooperation in the chat anyway, so I don't think it was too bad. In retrospect what I would have changed would be to make the third day hints, which gave rather explicit instructions, more vague instead.


Metapuzzle design
Due to this design, the first thing which had to be determined is of course the metapuzzle answer so that the regular answers would fit. I knew from the start that I wanted to encode a tech tree as the structure for the metapuzzle, so I started by drawing out the tech trees of each race for both BW and WoL, and seeing how many answers would be admitted and what the tree would look like.


The first to be eliminated was BW Protoss, as it has the unique and annoying characteristic that the Arbiter Tribunal requires both Templar Archives and Stargate. The tree for WoL Protoss also did not look very balanced (Twilight Council leading to both Dark Shrine and Templar Archives, but other tier 3 buildings with only one follow-on). Zerg on the other hand I thought was difficult to encode the evolution of the Lair and Hive.

This left Terran. WoL unfortunately only really admits eight buildings, since add-ons are universal, which I felt was too few. On the other hand, BW Terran had a very nice looking tech tree which could be easily broken up into three sets (production buildings, research buildings and add-ons) with each set starting at a level higher than the others. This I felt led to a very nice encoding and gives 12 levels, which I felt was a nice number.


The next step was to find a phrase for the final solution. I tried a number of different player IDs, names, teams, units, maps, etc. until I settled on Sniper's Ridge[sic], a map played in the last season of Brood War. Unfortunately the canonical name of the map is actually Sniper(no s) Ridge, which is only 11 characters. By that time I had already started filling in possible solutions for the regular stages, so I just kept one of the levels for a final trivia question at the end.


The storyline
This was possibly the worst thought out part of the design. I intended there to be a story, for several reasons: something to enjoy if the puzzles were too hard, and as a means to subtly hint at puzzle answers with keywords in the story.

However, as a not terribly creative person I couldn't really decide on a direction and did several rewrites of just the introduction. I had hoped to include more references to notable characters around SC2SEA, but was unable to really fit them in, so that idea got dropped. The pacing of the story is incredibly bad, it doesn't make a whole lot of sense, and I never got around to giving it a proper ending. If I were to do this again, I'd probably engage the help of someone more creative than me to help craft at least the plot of the story.


Stage 1.0
Answer: dustinbrowder
Common incorrect guesses: sudowoodo (most by far), rock, brock, destructiblerocks, squirtle, pokemon, scv, drone

Originally the newspaper clipping was intended to be part of the actual story (DB was being held hostage, and you would meet David Kim on your adventure. However I wasn't able to write that in such a way without revealing the answer very quickly, so it was turned into a standalone thing.

Identification of the character was intended to be relatively straightforward, and I had hoped the cue to types in the puzzle text would quickly lead people to Rock, but that didn't quite happen. I'm guessing that Sudowoodo blocking your way in the game possibly led some people to guess destructiblerocks.


Stage 2.0
Answer: freneticarray
Common incorrect guesses: fray, fennerteam, carnage

This was a relatively straightforward puzzle to design, and basically came down to what I could spell out using element symbols. Originally it was designed as FReNeTiCArRaY, but later changed to FeNNErTeAm just to require a little knowledge of SEA teams. The most common incorrect guess was the tag version fray which is not too bad.


Stage 3.0
Answer: flash
Common incorrect guesses: 1kzj6wdJeyQ

This was also a puzzle designed very early on, and was just an encoding of a string into go to it and determine the significance.


Stage 4.0
Answer: drone
Common incorrect guesses: terran, nomatterthecost, echoesofwar

The answer to this was one of the last to be set, but the format of the puzzle I decided upon extremely early. Echoes of War is a Blizzard image album which was recorded in Australia by Eminence Symphony Orchestra.

As Pox quickly determined, this was inspired by a MUMS puzzle, 3.4 from 2012. One of the things I hated most from MUMS puzzles was the need to identify a lot of unknowns like pictures or songs, so I made this easier by sourcing all the tracks from one album and making them all StarCraft-themed.

The file provided was multi-channel to allow listening to each track separately. Unfortunately, the order of channels isn't well defined, so some anagramming was necessary.

Originally, the word spelled would be album and the answer was echoesofwar. However, I felt that was a bit backward, as you would know that already, so it was rearranged to give drone.

This had the lowest solve rate on the first day, mostly due to difficulty in identifying the proper names of the tracks (Terran Theme 1 was thrown around a lot), but once people realised Echoes referred to Echoes of War the solution appeared to follow relatively quickly.

One problem appeared to be some bad mixing between the tracks, because some were too quiet to be properly heard in a stereo mix.


Stage 5.0
Answer: corvidreactor
Common incorrect guesses: energyupgrade, theraven, raven, various energy upgrades

This was one of the first answers and titles I filled in, because it was just too convenient to use the classic poem The Raven. Pretty straightforward to design, the most difficult part was finding stanzas with the appropriate letters available.


Stage 2.1
Answer: crackling
Common incorrect guesses: infinity (?)

All of day 2 was intended to be sort of a free-ride, where the puzzles just involved working out the trick, and then the answer was sort of just presented.

At one point I had started to insert alternative answers in the grid based on other orderings, but then decided against it and just filled it uniformly with random letters. The story originally hinted more strongly at the answer, by having you look at page 3-3 with adrenaline pumping, but I felt that might be too much of a give-away so dropped it to just a reference to page 33.

This had an error when it was first published, which I thought I had corrected, but as it turned out I had corrected my local copy and then 'published' it the wrong way and overwrote the correction. ><


Stage 3.1
Answer: marauder
Common incorrect guesses: mvp, stalker, random, various Random players

Oh boy. This one, which was meant to be a five second thing, turned out to cause a lot of confusion and angst, and I consider one of the worst designed puzzles of the set. The answer to this one came when I was filling in possibles by just asking my friends to give me StarCraft related words and phrases. I wrote down marauder, and it just ended up sticking.

This was originally going to be a WebGL puzzle, with a 3D interactive dice and some game of chance involving dice which you would play. However, I ditched that idea as it ended up taking too long to do (I never got round to designing an appropriate game, but did have a rudimentary dice thing up and running). So, I instead just stuck the texture of the dice up in flat form.

The clue about MVP not getting five was supposed to point to something wrong with position 5, but it seemed to point people in completely the wrong direction. People also seemed to get hung up on the stalker being wrong without also considering it's pair, the marauder. I'm not sure why this was.


Stage 4.1
Answer: reality
Common incorrect guesses: a massive spam list of le?t?iy

This was another puzzle inspired (or in this case, pretty much completely ripped off) from MUMS, in this case 3.1 from 2007. Originally this was intended to be a story-only puzzle, where the key words were to point you to Miles' Whos' Who profile and streaming.

However, those hints never really stood out, and with the story being in flux for so long, I last minute added the clocks puzzle instead, with plenty of reference to where it was. The ordering being important appeared to be missed by many.

There was also originally supposed to be a reference to being sued too much (Reality being a player on Samsung Khan) but when I wrote the final story for this it didn't have a natural place to fit so was dropped.


Stage 3.a
Answer: cloakingfield
Common incorrect guesses: mothership, darktemplar, observer, light, cloaking, various passive abilties

The only remaining story-only puzzle. For some reason people assumed the white blob was a light, when it was supposed to just have the effect of making things underneath it invisible (hence cloaking field).

That there was a lot of guesses for mothership is a good sign, but the name of the ability seemed to elude people for a while, possibly because there was no real confirmation that mothership was the right track. If I were to do this again, I would possibly add a random gif that popped up from time to time to represent a vortex or something to confirm this.


Stage 4.a
Answer: xgking
Common incorrect guesses: delete, broodwar, 266920, xeriagaming, infestor, Tim Twist or Trafalgar Law (?)

This puzzle (or at least the first step) I feel is very crappily designed. There is very little indication to look at the hex values of the colours, and no real reason there are there. More or less, when I designed that part I had run out of ideas to encode a value. The replay portion of the game was decided early on in order to fit the Control Tower stage.

The game itself I played with my friend (VBsirhcle) the night before the puzzle went out, basically to ensure it wouldn't leak out. It was actually the third such game, and yet I still didn't leave enough room to fit the Creep Tumor writing (because I kept making the letters bigger than I planned). It was also difficult finding a sufficiently flat map with buildable areas in order to do the writing.

It was a bit unfortunate that the answer for this one was the tag (xGKing) when an earlier puzzle had the full name instead of the tag (Frenetic Array), so the word TAG was added to clarify this.

I'm not sure where the "Tim Twist or Trafalgar Law" guesses came from...


Stage 5.a
Answer: sktelecomt1
Common incorrect guesses: sktelecom, skt, sk

Once again, a relatively straightforward one, the answer was picked very early on, and the telephone keypad was related to the answer and so was picked. The first round clue may have been a little too explicit by saying "Hertz" (originally planned to say hurts instead, and changed at the last moment).

More or less picked because SKT is best KT


Stage 5.b
Answer: pig
Common incorrect guesses: proannn, mafia, rossi, light, various players, various "community open ??"

The original answer for this was going to be yoonyj, the winner of the most Community Opens at this time and the puzzle was story-only. Unfortunately, as I decided against these, adding a proper puzzle became difficult (as I could pull the same trick as the final puzzle with YYJ but it was hard to also point to requiring the full ID, and the letter it contributed was N). So in the end, the answer was switched to PiG, and the letter N was switched to another answer.

The picking of names was slightly difficult. Ideally I wanted all finals appearances to count, but there were too many combinations of players that picking a subset was too hard. Instead, I finally narrowed it down to winners and their respective runners-up, and then carefully excluding further candidate players and filling in the rest with randoms.

It is slightly unfortunate that the answer could be stumbled upon, as I don't think many people actually obtained it by the intended route. It seems that identifying it was CO-related was easy but then the next step was very unclear. It is also unfortunate that some winners with no respective runner-up had to be included due to a limitation of available player IDs.



Meta
Answer: sniperridge
Common incorrect guesses: fusioncore, samsung, heartoftheswarm

The final meta stage was delivered in the form of a simple game of snake delivering the chunks of the puzzle string via the colour, text and point value of each pick-up. This I felt would quite a fun way to do so, having dropped the only other interactive puzzle because it was too time consuming. Snake is a very easy game both to implement and play, and so was ideal for the job.

Given that the puzzle data had to somehow be delivered to the client, it was convenient in the end that there was a prime number of letters, allowing them to be stored in a mod 11 permutation easily. I felt if you were willing to untangle it yourself from the source, that was a valid solution method and no harder than playing snake.

The wall was a late addition (implemented the morning it went out), just to make it a little harder, and also hint at the fact it was Terran-related. (Actually, it was also originally not wrap-around, so the borders killed you, but I changed that early on as it meant I had to keep moving while testing.)

The story was intentionally made extremely short so that the two hints were not lost (brooding -> Brood War, tree -> tech tree), but it appears that didn't work, and people were still lost as to how to proceed.

I am glad that people identified the colours to be blue and violet almost immediately (I had intended to use HTML colour violet, but that looks more pink than anything else), but the connection to the building and advanced building hotkeys wasn't as forthcoming as I had hoped.

It seems using a Brood War tech tree was a bit evil, as when the final super hint gave away this fact, there was a lot of "WTF!" going on (fortunatenly Pox had already deduced this by that point, so it turns out the super hint was unnecessary). One problem was that people were taking "sc" and "sc2" to be synonymous, whereas I was very careful to distinguish between them when asked about whether answers were "sc-related" or "sc2-related" in chat.

The somewhat strange stage numbering scheme also seemed to work, with people questioning what it meant, although it being a tree structure seems to have been less apparent.


Final thoughts

This was a lot of fun to put together, but it also took way longer than I thought. Next time, I'm definitely finding someone to spread the work out! Many puzzles ended up in the format 'some encoding of the puzzle string' which decoded into a 'hint at the answer' to give the final solution. There were a lot of ideas which got tossed out because they were too time-consuming, or didn't fit nicely.

I tried to make all the puzzles such that they didn't require blind guessing or excessive identification of unknowns and grinding to get to an answer. They were also designed so that there was a 'pit of success', so that you could tell you were on the right track. With the exception of a couple, I think that was mostly successful.

You can probably tell which were the ones I designed later, as the quality tended to drop as I got busier in real-life. Time constraints also meant I was unable to have someone test solve the puzzles to check their accuracy and difficulty.

If we had the capability to have more, or a divisible prize, I would have definitely liked to do this as a team contest like most other puzzle hunts around the world. Solving with others is always nicer to do (as I think most participants discovered), but unfortunately the nature of the prize meant that this wasn't to be.

I want to thank everyone who persisted with it, and I hope you found it rewarding when you managed to get the answers. It was intended to be challenging, but I am a little disappointed at the take-up, particularly during the later stages, although the real-life timing probably makes this an incredibly busy period for most people. There was also not quite as much abuse in the guess log as I had hoped I hope you had fun, and if you have any questions about it, please ask!

gg