Hunt Wrap-Up

20 teams (14 full ranking and 6 partial ranking) comprising 140 solvers took part in SGPH2025, solving 221 puzzles in total. Top props once again to Eggplant Parms, who with just 4 solvers kept up a good pace and unlocked all three House metas by 8.21pm. They got held up by Sonic Tank for some time before being the first team to finish the hunt at 10.09pm, pretty close to their finish time last year. Big congrats too to HWNK, White Maria, SIMOWO and Hee-ho, who also finished the hunt on Sunday in this order. Well done too to 3 other teams - Phoenix, RSVPuzzle, and Yuki, who also managed to uncover the House secret weapons and solved at least one House meta. Great job and thank you to all our teams and solvers for being part of SGPH2025!

Below is the interactive team solve progress chart. You can select/deselect the teams to be shown from the legend, and also choose the time period to zoom in on for the x-axis, using the sliders at the top of the chart.

First correct and fastest answer submission - By Eggplant Parms for Trike after 11min 46s from the start of the hunt.

Last-grasp correct answer submissions - Phoenix solved and submitted the correct answer for the Sonic Tank meta just 19 seconds before the hunt closed on Sun 11:59pm! This was the last correct answer submission in the hunt. RSVPuzzle similarly solved Death Hand just a little over 2 minutes before hunt closed on Sun 11:57pm.

Highest submission accuracy - Hee-ho solved 21 puzzles with 14 incorrect submissions.

Amusing incorrect answer submissions - Phoenix sent in a number of funny guesses from extrapolation of partial letters, such as NONCASUALTYOFVOICE for Palace, SCRAMBLEDINTOSPICEDCRATERS for Refinery, CHILLIPADI for Refinery, and OUTRIGGERCANOEING for Windtrap. SIMOWO also submitted PADDLINGAVAN for Windtrap.

Commendations for Saturday overnight solvers - As usual, a number of teams continued solving the puzzles overnight Saturday. Most notably, Yuki solved Death Hand at 12.05am, unlocking the meta Devastator which they solved at 12.15am, and then solved Starport at 2.11am. SIMOWO solved Saboteur at 1.20am, and resolve solved Trike at 12.24am. Phoenix also solved Palace at 12.40am and Refinery at 1.01am. Many other teams like ._., The Poos, Fun Run, HWNK and White Maria were active till late as well, and some had solved puzzles earlier before midnight. Great dedication and solving by everyone!

Most interesting hint request - HWNK sent us the most entertaining hint request in the form of a detailed ChatGPT prompt, well done! I was really surprised they took the time to come up with this late in the hunt – turns out it was something they had planned to do but only remembered then. I wanted to reply to them in a similar manner akin to a ChatGPT response, but the nudge for them was a pretty straightforward one-liner, so there wasn't an opportunity there. But perhaps next time!

Best task submissions - For Construction Yard, teams who sent in the extracted phrase from the puzzle received a task instruction to build and submit their model of a Dune II unit or structure (using any materials). As usual, the submission task was our opportunity for teams to exercise their creativity and to engage in something fun besides hunt puzzle solving, and is a common feature in the MIT Mystery Hunt as well. Three teams sent in their wonderful submissions below:

Hee-ho sent in this amazing model of the Harkonnen Death Hand, after being the first team to get the task submission instruction. I had not expected teams to replicate the unit images exactly, but that's what this was (as seen from the inset image), and it was epic!

RSVPuzzle also followed the Sonic Tank unit image (seen in inset) closely when making their model, on some Lego builder app it seems. Another awesome submission, with the camouflage pattern of the Sonic Tank really on point!

._. received the task submission instruction less than 7 minutes before hunt closed, but managed to send in this model of an Ornithopter in time. Unfortunately, we did not manage to respond to their submission before hunt closed, but nonetheless, a great effort on this model and solving the puzzle!

Solving hacks by teams - For Spice Silo, Moo's Clues sorted the clues in alphabetical order for fun, and noticed that they were again colored alternating red and black, which seemed too much of a coincidence. So they rightly followed their instinct and played out both games this way, inadvertently skipping ahead to the next step without noticing the clue phrases from the first game. But due to this, they also tried incorporating the unused letters from the first game and got stuck at the last step for quite long.

For Construction Yard, rather than use physical models, HWNK simply built all the required models inside Minecraft instead! They were really proud of their solving hack too, and shared what they did with us. And I think these look really great as well!

Last year's hunt had only one final meta (which made use of other answers), although not many solvers realized that the 6 milestone puzzles (the Guardians) that gated the final meta, also behaved like metas in this regard. Metas are a key trait of puzzle hunts, as they culminate the solving of all the other puzzles (typically in a hunt round) with a self-referential puzzle that uses these other feeder puzzle answers to solve. So there was feedback from some solvers that they would have preferred more meta milestones.

However, metas are by their nature tricky for solvers. They typically gate progress to unlock succeeding rounds (hence are milestones/finales). In past SGPHs that had 2 or even 3 consecutive-round structures with 2-3 metas, the majority of teams were held up at the first meta, and did not get to spend much time solving the rest of the hunt/puzzles. So I employed the parallel round structure this year (last used only once in SGPH 2019) to try and give solvers the chance to unlock and solve more metas, but without these metas gating any progress except to the final meta. To this end, the metas this year were also made intentionally short and straightforward, as I wanted solvers to be able to clear these milestones without too much trouble once they had met the unlock requirement. Even the aha for the final meta was relatively easy, though this was also partly because it used the 3 pun answers from the metas, and it typically isn't trivial to construct a second layer of commonality and pun into these.

The intent of having a 3-parallel-round structure, and providing 3 metas, was in hope that teams would strategize and focus on certain rounds/puzzles together first, in order to unlock a meta to solve. Especially for teams who are not able to complete the whole hunt, they could view the 3 rounds as akin to 3 separate mini-hunts with a meta milestone achievement point. But perhaps it was foreseeable that solvers would tend to just try to solve any puzzle they liked or looked shorter/easier instead. So it was unsurprising during the hunt to see (and confirmed later from the feedback) that teams did not actively target any particular rounds/puzzles to unlock metas, even towards the end of the hunt. Similar behavior was observed even after metas were unlocked. Teams that were stuck on a meta often continued trying to solve other puzzles, instead of focusing resources together on cracking the meta first.

Having more rounds/metas could also set up undue expectations. Solvers tend to be more familiar with encountering metas that unlock early with fewer feeders solved. Teams could solve the meta with partial feeders, and have choices in the feeder puzzles they need to solve. But to solve the same number of puzzles, this approach requires 2-3 extra puzzles per round/meta to be written. We are a small organizing team. I write most of the puzzles and the process is very time-consuming, so it would be challenging to always have more extra buffer puzzles to provide this option of choice. I emphasized a lot that SGPH is intended for 8 solvers, and teams should fill out their roster to make more progress. One key reason is most solvers have a preference (and limitation) on the type of puzzles they solve. And puzzle hunts normally have a wide variety of puzzles, so you can expect to only care for a fraction of the puzzles in a hunt. Therefore, having more team members not just means having more solvers, and also with a mix of puzzle interests and skills to complement each other and help solve the other puzzles that you are less inclined towards in a hunt. Overall on review, the recent SGPHs which had a single round/meta structure seemed to have worked better for our solvers, and is likely to continue to be the approach unless the hunt theme requires otherwise.

This year, the planned hunt structure of 3-parallel-rounds drove the conceptualization of a suitable theme. We spent quite a while initially brainstorming on a different theme, trying to make it work but without success. Eventually, I switched to the Dune theme, as the sequel film was just released in 2024, and would likely be something that solvers were broadly familiar with. Most importantly, there was a very fitting narrative using the setup of the classic 1992 Dune II videogame by Westwood Studios. Dune II was the founding game in the real-time strategy (RTS) genre, and some solvers should also be familiar with it. The 3-parallel-round structure maps to the 3 Houses, with their in-game units and structures providing plenty of material for individual puzzle themes. The metas mapped to the unique special units of each of the 3 Houses, with the final meta being the iconic Sandworm. Like last year, the videogame theme also lent well to the puzzle imagemap, which incorporated the "fog of war" hallmark of RTS games.

Work on the theme and metas started early around October of the previous year, and took longer till December to finalize because of the change. The puzzles took around the next 6 months as usual to conceptualize, write and test. During the construction, it was evident that this year's hunt would inadvertently be more challenging than last year's, and there would likely be fewer solves and teams that complete the hunt. This is because last year's hunt had a number of puzzles which were relatively straightforward, like filling a grid with clue answers (which solvers shared that they wanted more variety) and the aha was often just a grid-based extraction which has less possibilities. Conversely, many of this year's puzzles had ahas which are more alike those of MH puzzles, such as external references, and ahas before the extraction step (which had more possibilities and required more intuitive solving). The three logic puzzles this year also involved game logic, a possibly less familiar sub-genre compared to Nikoli-style logic puzzles. But perhaps not a bad thing that the puzzles this year are just a little closer representation of hunt puzzles that you might encounter in other puzzle hunts, especially the MIT Mystery Hunt, and the experience will hopefully help you become better solvers in future hunts.

So teams should not be too discouraged by the number of puzzles solved, or if you did not manage to complete the hunt in time this year. SGPH has a relatively short duration, so teams which have more experienced solvers will likely be able to do better. In the feedback, one team shared setting a target of solving a puzzle within the first two hours of hunt this year, which they were glad to have met, congrats! I felt it was a tad conservative though, and look forward to teams setting a higher target of solving a puzzle say within the first hour instead, and achieving that. The difference between a team spending a few hours to solve a hunt puzzle, and another spending a few minutes to solve the same one, is really mainly just down to experience and knowing what to do. A hunt puzzle is like a coded task. Even though it has no instructions, an experienced solver who knows the hunt puzzle conventions or has seen similar puzzles before can "decode" it quickly and knows intuitively what the puzzle requires the solver to do. Conversely, less experienced solvers might spend time not knowing where to start or even chasing different theories unnecessarily, thus not actually making any solving progress despite the time spent on the puzzle.

As an illustration, here are 3 steps that an experienced solver might approach the puzzle Trike (which every team solved) differently or more efficiently to solve in less time:
Triage - Which of the initial puzzles does the team want to focus on first? It's natural for some solvers to split off to work individually on a puzzle they like, but for others or newbies, it's helpful to triage the open puzzles to assess which one could be easier for a group of solvers to tackle together first. Trike, with its list of specific word clues and other observable traits mentioned below, should stand out more.
Analyze - How is each piece of puzzle info used? Map out what goes where, counting the number of items usually helps to determine what goes with what. Looking ahead helps you figure out what you need now, and potentially skip through most of the puzzle. For Trike, what is likely used for the final ordering? There is a given list of images at the end of the puzzle. And what is likely used for the final extraction? Each of the icons at the end have an associated index.
Solve - What is the aha that will help you? Look for thematic clues/connections in the puzzle title, flavortext and presentation, which together with some initial rotated clue answers, suggest the commonality of trios. And since you had already figured out ahead the final extraction method and ordering, you actually just need to solve/match less than half of these trios to possibly extrapolate the final answer (which is what some teams did).

Celestine as usual is our tech chief, without whom the hunt website would not have been possible. This year, we added a button to each page to allow solvers to contact HQ more easily (for non-hint purposes). Otherwise, the website was mostly unchanged from previous years... except, a couple of hours into the hunt, we suddenly realized that the server was unable to send automated emails and email sending was encountering "connection timeout" errors to the SMTP server. After some digging, we eventually realized that the server's hosting provider, DigitalOcean, had started blocked outgoing SMTP ports this year, probably as a means to restrict their instances being used to send spam mail. Email sending wasn't tested before this year's hunt as it had worked fine in previous years, somewhat of an oversight on our part. So we manually contacted teams to alert them to check back on their hint responses via the website in the meantime instead. For the next few hours we tried to find an alternative mail provider; fortunately, the volume of hint requests was still fairly low at this time. Right before 8pm Saturday (as hint volumes were starting to ramp up), we managed to find an alternative Mailgun, and switched to it for mail sending for hint response alerts the rest of the hunt, and thankfully, it held up. Automated emails are fairly important as alerts for our workflow, and given that more hosting providers are starting to implement spam restrictions, in future years this is something that we will have to look at more closely to avoid nasty surprises.

This year, we answered a total of 195 hints. Excluding the 5 hours or so overnight that HQ shut down, this meant around an average of 6.5 hint responses were sent out per hour! Naturally these rates were significantly higher around the timings when extra hints were released, and towards the end of the hunt.

Most hint requests tend to focus on asking for help to crack the handful of remaining clues that the team is unable to solve. This is not so useful, as puzzles can usually be solved without all the answers (by extrapolating any missing letters from the extraction). So I often reply to tell teams not to be too worried about the remaining clue answers, but to instead look ahead and try to figure out the next step first. Verifying for a team whether clue answers are correct could be helpful. But when stuck, the first thing to attempt should be doing a thorough re-checking of all the clue answers and steps already. Generally it's safe to assume there were earlier errors mades and try to find and correct these. Mistakes could also be caught at the extraction stage by fixing letters based on thematic or likely words. There are also teams that do not share the link to their spreadsheet in their hint request, and others that only share the link to their spreadsheet. But of these approaches are not ideal as we cannot tell if mistakes had been made somewhere, or figure out where exactly the team is stuck at without further details on theories and what has been attempted.

This year, I observed many teams were able to explain their theories on the likely next steps, which was great! Even if some of these theories could be incorrect, this practice of always looking ahead and processing where each piece of puzzle information intuitively fits is a critical solving habit that everyone should ingrain in themselves to be better solvers. In cases where teams have just missed something in their working or needed a minor nudge, I have sometimes refunded hint requests, as the team had done the right steps already and shouldn't remain stuck just to save a hint. After spending time solving a puzzle and making quite a bit of progress, it is much better to close it out with a hint or two if needed, as a solve is always more satisfying (get to unlock more puzzles too), and you would learn/remember more from the experience than if the puzzle had remained unsolved. As usual in my hint responses, I try to share any relevant tips or common hunt puzzle conventions, to help teams in tackling similar conundrums in future puzzles they encounter. You can also read this puzzle hunt guide for more of these conventions and solving tips.

Amusing spreadsheet from one of the teams reflecting their mental state after trying various approaches without success to solve one of the metas.

I also sometimes hung around in a team's spreadsheet to see how they solved a puzzle, and one useful practice seen is discussing the solving in the sheet itself. Typically, teams would have a separate Discord text/voice channel, which is important for overall coordination on which puzzles to solve together. But when solving a puzzle, I too find it very helpful to just type in the sheet to chat quickly with the other solvers. I find that this encourages everyone to put in their observations and ideas on the sheet, which gets recorded and seen by others, as opposed to voicing out something which might get lost amongst a voice chat. Solvers (especially introverts) are also more likely to contribute their observations and ideas onto the sheet directly. Perhaps this is something that teams might want to consider trying out next time, but still important to keep the sheet tidy to avoid meaningful ideas being lost. Also, it's always great watching live the moments when teams realize the aha and cracked the puzzle together - I believe these are the memories that make solvers enjoy solving hunt puzzles.

Do follow up this momentum and take part in other upcoming online puzzle hunts this year together with your team, as the solving experience really helps! Join us at the SG Puzzlers Discord group (and the SG Puzzlers Facebook group) to keep in touch, be informed of good upcoming hunts, and find members for your hunt team. Feel free to invite us to join you to solve a hunt too! We look forward to seeing you all in the Discord channel, and hopefully at other online puzzle hunts too. We are also opening for limited applications to join the Puzzlesmiths team to help organize the annual SGPH. If you have a strong interest in writing puzzles, and are interested to contribute, and improve through a proper puzzle and hunt writing process, email and tell us a bit about yourself! We also welcome those with skills in tech and/or art.

Another recommendation is to go solve some past puzzle hunts (ideally with others in your team) for fun and practice! There is a good supply of archived hunts with ready puzzles available online, so you don't need wait for a live hunt with new puzzles, or write them yourselves. I strongly encourage trying puzzles from past MIT Mystery Hunts, which generally have well-tested puzzles that follow standard hunt puzzle conventions. This allows you to solve in an intuitive and logical manner, giving you a better hunt puzzle foundation and reinforcing your solving skills. Recent Mystery Hunts usually have an initial round with many puzzles that cater to solvers of various experience levels. But also helpful to try solving some of the Mystery Hunt puzzles from the main rounds, which would be a good experience of the typical kind of hunt puzzles that could be expected, as well as the general difficulty of standard hunt puzzles. There are also many other archived puzzle hunts out there you could practice solving, including past Singapore Puzzle Hunts.

Feel free to post any other questions or discussions on SGPH in the Discord channel. I will be posting about some recommended upcoming online puzzle hunts there soon, keep a look out for this. Do try to take part and solve these hunts in your teams as well, especially the MIT Mystery Hunt next Jan which should be the milestone hunt in everyone's annual hunt calendar. Besides the other online puzzle hunts, I have plans for two other events that you could look forward to this year:
- A year-end in-person event before the Mystery Hunt, this would be an informal opportunity for local solvers, including those planning to solve the Mystery Hunt, to catch up before the hunt.
- A hunt puzzle solving and writing workshop, as part of continuing to help grow and give back to our local puzzle hunt community.

Please email us at puzzlesmiths@sgpuzzlehunt.com or message to let us know your thoughts on these - Would you be interested to join us for these, and what kind of activities would you like to see? And affordable venue is always a challenge for these in-person events, so any good venue suggestions or opportunities would be much appreciated too, thanks!
PS. Some of you might have encountered issues with sending us emails this week, this has been fixed. So feel free to send/resend to our email above, or puzzlesmiths@gmail.com.
More details will be shared in the Discord group when plans are finalized, so remember to set your message notifications for the server if you are not a frequent Discord user to avoid missing on updates.

So hope to see everyone again before the big hunt of the year! And stay tuned for SGPH 2026!

Hunt Editor
Ong Kah Kien
on behalf of Celestine Lau and Sim Hong Jhun (Puzzlesmiths)