[002] Despair (Shuffled World Devlog)

Follow updates on my Substack: https://melodicambient.substack.com/

Refresher & The Problem

Well, it's another weekend, so I'm back to thinking about my side project a little. Today I'm taking an experimental approach which is to write at length until I come up with some solution - sort of like talking to myself.

Here's the first post which contextualizes things a bit. https://han-tani.itch.io/shuffled-world-codename/devlog/454613/codename-shuffled...

 In summary, the key things to know about Shuffled World are:

  • 3D. No combat. Story-focused. Treasure-hunting game. 
  • You walk around at a fixed camera angle and can jump but it's not a platformer in focus.
  • The world can be reset. When it resets you start in a Home Room.
  • Each Room - (a tiny level, not a literal room) has four doors leading out of it in the 4 compass directions.
  • By opening a door, it connects a totally random room to the current room. By using a key, you can influence the door that gets connected.
  • The rooms are hand-designed by me. so it's not procedural like in a roguelike.
  • The world can be built into any shape.
  • The world is represented as a 2D grid, like a zelda minimap.
  • Certain arrangements of rooms make it possible to use a Treasure Key

Right now I'm dealing with a design problem, and the issue is that I can't really see good steps to prototyping my way out of it. (Spoiler by the end of the post I figure something out) 

A lot of my narrative theming, and what I want to write about (Cultural exchange, people picking different life paths/interests, etc), hinges on the idea that the world's 'normal places' are all isolated from each other, and the only way to reach another area is to reset the world, and build out a map that will let you reach another 'normal place'.

That is, Town A and B can NOT be connected by a regular RPG world map: what's important to the themes (around cultures, etc) is that it's nearly impossible to get between the towns without prior knowledge. It's not that there's a mysterious maze between towns (like the Lost Woods of Zelda), but it's more that it's impossible to really feel an orientation between Town A and Town B. In a regular game you might know "oh, my home is south of the castle!" but in Shuffled World, it needs to feel like "to reach the castle, I need to make the world into this configuration and hope for the best..."

It's tricky because I of course want the player to make their way to the 'normal places'. But merely describing "oh you can reach the castle by creating three successive Forest Rooms" also feels perfunctory: there's nothing interesting in finding out the solution. It's still possible to make it fun to execute the solution, but it feels less like testing a hypothesis than following a pre-ordained map, which kind of contradicts the narrative tone.

Of course, I could take an approach like in All Our Asias - where there simply is almost no gameplay at all except walking. But I feel like that worked for AOA because the story is about going deeper and deeper into the Memory World, following a story that is directly related to the deeper areas of the world. Whereas in Shuffled World, the narrative only works if the towns and treasures of the world truly feel isolated by this semi-random world you have to build out.

When I'm stuck...

When I'm really, really stuck in game design, usually I discuss with someone (usually Marina if it's one of our games) or ask for help on Twitter or Discord or forums. Sometimes I'll research a bunch of games to see some glimmer of my problem at another angle. Sometimes I'll read articles or books to try and gain insight into other designers' way of thinking.

 Of course, sometimes being stuck is a matter of needing to prototype more. You can't tell if an moveset design is fun until you're moving around test levels with it in the engine! I think this is something that's tricky to figure out without finishing games, but I think in my case I have a research/design problem right now, not a prototyping problem. I already prototyped what you see in the video above, which led me to the current problem. I think I need to think a bit about what I need to prototype next, rather than coding more and hoping something clicks.

Shuffled World seems to take influence from roguelikes, puzzle and mystery games. Those genres are both not my design forte nor my primary interest as a player. Despite that, I can appreciate an expertly-designed game in those genres, and there are takeaways to be had.

Derek Yu's book on Spelunky makes an interesting note about the "Chain" (one of Spelunky's greatest challenges) and the games' procedural nature : "The permanence of the Chain and the randomization of the world work in tandem to give you a greater appreciation of the other." This got me thinking a bit about how I should be balancing the random and fixed elements of my game for maximum effect. Of course, Shuffled World is not a roguelike nor would I describe it as procedurally generated, BUT it's still very helpful to see another designer talking about a balance between permanence and randomization.

Puzzles and Mysteries

After reading some interviews with puzzle designers like Alan Hazelden and Jonathan Blow, it reminded me of some of the base principles of making satisfying puzzles. What's interesting is that the principles they outline don't just apply to purely puzzle games, I think they're useful (and have analogues) in stuff like platformer or action level design, too!

The ones that stand out to me are: 

1. It's quick to test solutions. It was great to see someone talk about this, as it reminded me of the main problem (in my head) of my proposed solution for exploring the Shuffled World: having to walk around in 3D and try out doors for a while to know if the world is in the right state to spawn a treasure room or destination is a very long way of testing a solution.

And yet, the problem is that walking around in 3D is integral to the storytelling and sense of place in Shuffled World. Uh oh! 

2. Good puzzles are built on legible axioms. In other words, "you learn rules as you play".  In a puzzle game like The Witness, you might learn that black and white dots need to be divided. In a platforming game like Sephonie, you might start to get a sense of where, spatially, you can reach when you possess a high-jump-pod. When the game starts to throw different rules together, things get fun as you have to try and figure out how to satisfy multiple things at once. In an action game it might mean how to safely navigate a room with multiple enemy types, in a platformer it might mean trying to figure out how this room with two types of obstacle needs to be navigated, in The Witness it might mean how to satisfy a puzzle with black and white dots and some other puzzle mechanic.

I also researched a tiny tiny bit into Mystery games, which reminded me that these games tend to be more fun when the solution isn't obviously laid out for you. A GMT video on mysteries makes use of games that require you to type in possible answers  (or fill in the blank in a way that prevents brute force) (Her Story, Obra Dinn, some older mystery games).

Basically what happens here is that you're forced to find the patterns amongst the clues yourself, before testing out a solution. There's a very simple version of this in Uki-uki Carnival where you might pick from a bank of search terms (Corn Soup, Melon Bread, etc) to try and find a page that might have info you need to progress the game.

Rethinking the 3D Shuffled World?

After thinking about this, the first thing that came to mind was building the world out of a simplified map interface. Instead of walking to doors and using keys, you just sit in a room and build the room out with the same system. Well, this sort of sucks, since you'd just walk through the pre-built world and that wouldn't be that interesting, huh! There's a lot of meaning you can do through a game entirely made of UI like this (take Ruina for example) but it's unsuitable for the visual presentation and spatial tone I have in mind.

But, you'd be able to quickly test out solutions for finding treasure! So there's definitely something useful here, even if it's unsuitable as a complete solution. Sometimes it's important to pretend to throw everything away, but not really do it... since you might see something in the (hypothetical) destruction.

Well, okay. Back to thinking about Shuffled World. Mechanically there's two steps to finding treasure (vaguely meaning: literally treasure, or like a town or landmark).

  1. Planning your Route: Figuring out the proper world state that will allow the Treasure to be spawned
  2. Executing:  building out the world so that you can findthe treasure. 

I think what would get frustrating is if Planning involves a lot of inefficient trial and error. So I feel that should be somehow framed as a traditional puzzle experience, where it's easy to try out answers.

Executing the route is where we get our spatial meaning. Walking through 3D, opening doors. This is what creates the sense of being truly in some 'world that's all shuffled up'. You see the same forests and ruins over and over, but somehow, you end up at your destination. You leave that destination, knowing that you probably won't be able to take the same path there again in the future...

The design challenge here is it could be perfunctory building the world out. Thus I think the 'solution' (of what a valid plan is) should be wide, and building the world properly should be flexible, with some chance (like a card game.), so it still feels a little like exploration.

A Bad Solution

Let's say we're trying to find the "Magic Ruby". The solution to finding a Magic Ruby is to create a "Large Desert" - a contiguous region of 4 Sand Rooms in the shuffled world.

 Here's what I think would be a bad way to get a player to do this.

- You talk to a NPC and they tell you "rubies tend to appear in large deserts!" What does that mean? Who knows! Definitely not the player! Maybe you find a clue of a map that says "large desert": it looks like a minimap of some version of the Shuffled world, and features 4 sand rooms close together. You don't know for sure, but maybe "large desert" mean a patch of 4 sand rooms. You pick your Keys, reset the world, and try to build a room with those rooms together. Maybe you get it slightly wrong, and you can't find the rubies. You just wasted 10 minutes, that sucks! You don't know where you went wrong! Maybe you got it right and you don't know why. Sure, you progressed, but it wasn't that satisfying... either way, it's kind of mushy and took forever to try.

Or another bad way:

- The NPC just tells you the answer. They banter with you as you walk and follow their instructions to finding the Ruby. Okay,so there are some story opportunities here, but overall it just feels a little boring. If an NPC is to accompany you, there are better things an NPC can spend time saying than merely what to do next.

A potential solution??? Step 1: Clue Maps

What if, instead, we are given four "Clue Maps" of shuffled worlds where other NPCs have successfully found - or NOT found - Magic Rubies? Through these clue maps, as well as some verbal clues ("patch of desert"), it should be possible for me to design the clue maps so that a player can deduce a pattern.

For example, we could have these clue maps, which try to illustrate that a "Large Desert" is 4 or more contiguous Sand Rooms, and to find hte Magic Ruby, you need to use the Treasure Key while inside the Large Desert.

1. An invalid map, with only an L-shaped patch of 3 Sand Rooms.

2. A valid map, with a 2x2-shaped patch of 4 Sand Rooms.

3. Another invalid map

4. Another valid map.

A player can look at these - and try to figure out "what it is that makes these maps valid but not these other ones?" It's like an SAT Math problem! I love it...

I think this puzzle system is potentially axiomatic. For example, what if the desert had to be far from the starting room? How would you convey that in clues? Well,  an early treasure's set of clues could communicate that distance from home can matter. Another treasure could convey what a "desert" is. Then a later treasure's clue maps could combine both: There's a invalid map of a "close desert", but a valid map of a far desert, etc...

So through these clue maps, the player can make a Hypothesis. How to test it? Map builder!

Step 2: Map Builder

After viewing the clue maps, the player can make a map test of their own, using a "Map Builder" (something like Etrian Odyssey or the new Link's Awakening I showed at the top of this post maybe.)

First, they pick which treasure they're looking for. In this case, it's the Magic Ruby.

The game then lets you reference the clue maps as well as show you a list of verbal clues or hints you've found.

Then, the player can build a minimap and test where they think treasure would be. If they're right then the game will tell them so! They can then save this minimap for reference when actually exploring in 3D.

While building the minimap, they can use various measurement tools, both on their in-progress map and the clue maps. For instance, you could measure how far a room is from home. You could measure how much % of the world is made up of desert rooms. Etc.

What's nice is because minimap building isn't limited by the randomization rules and Key system of the 3D exploration, you can quickly test out different solutions to try and figure out what, exactly, are the conditions for finding the Treasure.

Will it be hard?

Anyways, I don't think these puzzles will be brain teasers on the level of hard puzzle games, but the important thing is that they should be kinda fun and satisfying to puzzle out. And they should fit nicely into the narrative of piecing together old stories (perhaps clue maps are stories of people's travels!) in order to find certain treasures.

What I like about this is it's abstract enough to fit into a variety of stories. As a total random example what if there's a line of 4x1 rooms you can spawn once, called the "Black River"? Some Treasures involve using this Black River, and those Treasures's Clue Maps could be related to certain rooms being on one side of the Black River or not. And this could tie into story. (Maybe part of the solution is there are only Blue Towns on one side, and only Green Towns on the other).

To me this is where things get fun: at the bottom we have some basic pattern-deducing, but then I can try bringing that up to the level as something narratively relevant. Hooray!

Step 3: Building the World and Keys

Here comes the Keys System. I think each Treasure will feature a limit on the number of Keys you can bring in, like size limits on a deck of cards, some Keys are more powerful than others - and thus - more expensive to include in your Keyring.

So you can't simply recreate a minimap, as mentioned in earlier posts, and there's some importance in trying to make the simplest valid map you can.

The Keys you use on a door merely influence what kind of room you can get. Sometimes you can control it, but you can't just pack your Keyring full of 4 expensive Sand Room Keys to make a Large Desert easily.

I think Keys can function a bit like cards in a card game: there are ways to open doors in the world that can optimize your chances for certain results, just like there are strategic plays of cards that can optimize for certain results.

 For example, you might be able to use a "Sandstorm Key" to create a room that has the effect: all new rooms connected to this room have a greater chance of being Sand. All you have to do now is hold out for two Sand Rooms to be connected to each other, and then use your two Sand Keys and you'll be able to use your Treasure Key to find the Ruby!

Hopefully you can see why it's important for the treasure solution to be flexible, not rigid. A "large desert" being 4 or more contiguous Sand Rooms is fairly flexible. It could be a line of 4 rooms, a T-shape, a square, etc... The number of world layouts that satisfy this is infinite, and if you know the rules, you should be able to reach the treasure consistently, if you choose your Keys wisely.

Exploring the world should have a Puzzle Philosophy more like Sephonie's ONYX Linking, where there are rules, light puzzle elements, but it's generally fairly flexible to win (like an action game) if you have a plan in mind. (This is why I view Sephonie's ONYX Linking as a "turn based action puzzle"...) It's pretty easy to reach a win state in ONYX Links, but the objects on the game board will influence the path you take to  get there, and if you want to win quickly, you'll have to think more carefully and fast.

Once you have a valid map plan, exploring the Shuffled World should be fairly straightforward, maybe akin to crafting complex items in games like Astronoka (a seed-crafting game) or the Atelier series, where you're exploring a random pool of stuff and using certain techniques to suss out what you want. 

There's a bit of luck involved, but it's there to spice up exploration, expose you to some surprise rooms (which might lead to clues for future treasures!), keep you on your toes.


- This was a fun post to write because I literally did not have a next step in mind at the beginning of the post. Actually, I was in despair... But after thinking at length (something about the magic of a blog post...), I now have something totally prototypable - not to mention an actual set of puzzle rules that I can Actually Design Puzzles For. I've gone from a kind of vague set of rules that seemed unsatisfying to something that I think not only could be pretty fun, but also enhance the story plans I have!

- Also: I also have a topic of puzzle design to explore! I can now go look for puzzle games where you 'deduce and then test the answers', looking for design ideas that might be relevant or useful. 

I hope this was a fun look into what the game design process looks like. A push and pull, where we rely on talking or reading the thoughts of other designers to hopefully unstick us from a problem. And to think out loud. And the importance of playing and checking out other genres, even if they're not what you're generally attracted to!

I'll see you next time where hopefully I have some of this built out!


Log in with itch.io to leave a comment.

I think a way to make these puzzles a little less potentialy frustrating would be to make rooms interact in some ways, and a solution for a puzzle would be to arrange them in a certain way. An npc could hint that the entrance to water town is behind a waterfall, and to make a waterfall you would have to put a river room next to a mountain room and it would create a new waterfall room in an adjacent unexplored tile.

I think it would then be fun for a player to experiment with different room combinations without any destination in mind, just to see what unique rooms they can make. Maybe the player tries surrounding a dessert room with water and it creates a little tropical island. Then these unique rooms could have clues for the more harder puzzles, which would be more convoluted so a player doesnt just stumble into them by accident.

Maybe new rooms could also change old rooms, so for example if a careless player put a river room next to a valley room, it would flood it and prevent the player from traversing it normally.

Also I think the GROW games are an example of a puzzle where the challenge is to guess a specific sequence of choices using clues and some trial and error, which I feel is sort of like the puzzles you described.

I feel like getting very specific sugestions could be annoying, especially when you have narrative themes already in mind and have to marry them with the mechanics, or maybe you already though of this and decided its not the right solution, so feel free to ignore me and my ideas and ban me from ever buying any Melos Han Tani game forever if you feel that way.

I really like this solution you landed on. I very much enjoyed the level of difficulty of the ONYX Linking in Sephonie, so having a similar difficulty for this sounds great to me. The gentle music/sfx in the video is nice too!

You mention there's a bit of luck involved. If I understand correctly, that comes from the fact that the room you get is never 100% guaranteed — so you could do everything "correctly" but still get a bad room. Would you just have to reset the world at that point? Do you think resetting/retrying like this would be tedious when you know the "solution" and just have to get good RNG?

FWIW, the board game Black Sonata came to mind when you mentioned "puzzle games where you 'deduce and then test the answers'". This video does a pretty good job of efficiently getting the idea across (the guy in the video is kinda goofy sometimes though lol). The "test the answers" part happens about halfway through the video FYI. It's a deduction game with an interesting physical design. Some cards have holes in them that you lay over other cards. If you see a symbol through that hole, that means you've correctly deduced something. I also see some similarities to the "shuffled world" in how the Dark Lady moves around.

Anyways, this post was really interesting to read! It's helpful (and entertaining) seeing how you work through issues. I'll probably try this "write a blog post" method next time I'm mega stuck haha.

I'm imagining that constructing the 'solution worlds' is some balance of knowing when to pull back from the more RNG-y Keys (ones that merely increase the rate of certain rooms) and utilize the more guaranteed Keys (ones that spawn certain rooms). So theoretically, with the right Key loadout, you should have a pretty low chance of failure - part of making sure this is the case is the solutions being flexible (e.g. a desert being 4 contiguous Sand rooms, vs. an exact 2x2 pattern - allows for some failure with RNG. Likewise if a desert is 4 or *more*, vs. exactly 4).

That being said, I do feel like there's something off to the whole solution I outlined in the post. Namely, I wonder if I'm looking at the 3D section of the game wrong - I feel like there's little significance to the rooms themselves, if most of the time you're just playing a little card game with adding on the doors. There's probably some way to mitigate this (maybe adding some diversity to what you do in the shuffled world - not just trying to construct treasure paths. Maybe some more freeform exploration options), but it'll take some more thinking/testing!

And thanks for linking that game! It's interesting what can be accomplished through physical games in terms of randomness/deduction.