Puzzling Out Puzzle Games

I’ve been meaning to write this post for a while, mostly to crystalize some thoughts that I’ve had while playing puzzle games, particularly in terms of what I like from them. A lot of what’s here is drawing from my experience of my favorite puzzle games like The Witness, Baba is You, Snakebird, A Monster’s Expedition, and others.

There will be single puzzle spoilers for A Monster’s Expedition and Cosmic Express, but no spoilers about hidden mechanics for any game. Also some technique spoilers for Q Remastered and Tears of the Kingdom.

I should also caveat that I have played a lot of puzzle games but haven’t actually designed one, let alone programmed one … yet.

The Narrative Structure of a Good Puzzle

Yes, I do place puzzles within the framework of narrative structures, but I’m a writer, it’s what I do. A good puzzle, to me, has certain elements in play, and though I wouldn’t say that they all need to be there for a puzzle to work, I do think that at least some of them do.

A puzzle is, essentially, setup and payoff. The setup is the puzzle itself, while the payoff is the solution. As contrasted with how setup and payoff work within a story, in a puzzle the player wanders down a path that the puzzle creator has set for them. The nature of being a puzzle creator is to create the path for the player. This doesn’t mean guiding the player toward the solution, though that’s part of it, it means creating pathways of play that are enjoyable as the player tries their best to reach the solution.

I was going to try to do this without examples, but I think that would be too hard, so let me boot up a game from my library and demonstrate what I’m talking about.

This is from A Monster’s Expedition. It’s a very early and very easy puzzle, picked because it’s literally one of the first puzzles that I had to put any thought whatsoever into. The goal of the game is to push over trees to get logs, roll or flip logs, then make a bridge to the next island. I’m about to spoil part of this puzzle, mostly to talk about what goes on in a player’s head, so if you haven’t played the game before and want to do it on your own … go get the game and play it, I guess. It’s a very easy puzzle, almost at the level of tutorial puzzle (which we’ll get to in a bit), so I personally think even if you plan to play the game, knowing the solution here won’t take away much, but I know especially for puzzle games, some people are extremely spoiler averse.

One of the first things we learn in A Monster’s Expedition is that trees can be pushed over to make logs, and one of the second things we learn is that logs roll if pushed from the side, or flip if pushed from one of the ends. A flipped log will flip one tile, while a rolling log will keep rolling until it knocks into something or goes into the ocean. This is as much mechanical complexity as the game has at this point.

When we arrive at this island (the unit of puzzle in this game) we’re presented with our first option, which is which direction to push the tree over in, which will then determine where our log starts. Already, before we’ve even done that, we probably see the clear end goal, which is that the log wants to end up in a horizontal position in the one-square gap between the two islands, which will allow us to cross. We have a clear setup and a clear goal, our only job is to get that log into the correct position. For a game like this, this is the starting state and ending state, and what’s left is for the player to bridge. In narrative terms, it’s the central conflict.

So, to first steps, once the land has been surveyed: which direction will we push the tree? We look at the tree, and imagine all the places that the log would end up, and where it might go from there. If we’re a certain kind of player, we actually do push the tree over just to see, and maybe try some things that don’t work. If you push the tree right, it can only be rolled up and into the water, or down into the water. If you push it left, you can only roll it down into the water. If you push it down, you can then push it left into the water. If you push it up? Suddenly there are options! If you’ve played A Monster’s Expedition before, you probably didn’t even really think about it, but this winnowing down of possibility space is a really core component of good, well-designed puzzles. Once the player has figured out that first step, pushing the tree up, they have additional concrete information about what the ultimate solution will look like, and will never regress.

This whole puzzle more or less continues this way, though it’s less forced than before. We push the log right, where it’s stopped by the pedestal, then we have an option to push it up or down, and we can go through like that, eliminating bad possibilities as we go. Every step of the solve gets us more information and, crucially, takes us closer to the end. We’re following a path. Each minor little thing that brings us closer to the end goal feels nice and pleasant, and the solution isn’t a flash of insight, it’s a series of deductions and minor bits of information that add up.

My best guess is that most people solving this puzzle don’t verbalize or internalize their solve of it. It takes seconds to solve, probably, and some of the dead ends you never go down because you randomly picked the right way the first time, or because your brain was able to automatically eliminate the wrong turns without active thought. This is foundational puzzle game stuff, and entire games have been made with just this sort of thinking. There are a few things that arise from this though, and I think they’re interesting principles to point out.

• Conservation of Detail
  A puzzle shouldn’t have much, if any, extraneous stuff in it. The example puzzle could totally have had a second tree that wasn’t a part of the path toward the goal. Would this have improved the puzzle? No, probably not. The player would have to eliminate a possibility and get no progress toward completion in return. In comparison to a novel, it would be like introducing a character who wasn’t interesting and had nothing to do with the plot.
• Red Herrings Should be Interesting
  In narrative terms, a red herring is a possibility that gets tracked down and then eliminated. In puzzle games, these are super common, since a puzzle where every step is “forced” wouldn’t actually be a puzzle at all. A better definition of a red herring is an element of the puzzle where the player is enticed to follow a chain of thought or action which ends up not being correct. A good red herring gives you some vital information about the puzzle and usually you get a feeling of “why do I not simply … oh, that’s why”.
• Puzzles Should Have One Unique Solution
  I think there are a bunch of asterisks here, but the “one unique solution” element is mostly about that sense of driving toward a solution as though engaging in a narrative by going through different steps. If there are two solutions, then they need to have their own little narrative paths to them, which usually involves different goals.
• Non-repetition Principle
  No one wants to read the same story twice, and no one wants to re-solve the same puzzle. A puzzle is interesting because of the paths you go down when you solve it, so going down those same paths is boring.
• Variation Principle
  Contrarily, there’s an interesting thing that puzzles do where you’ll do one puzzle, then do a puzzle that’s almost but not quite the same. This gives rise to themes, and themes are good, especially if puzzles of the same theme are grouped together. It’s like having a hero whose villain is a reflection of them.
• Blocking Principle
  The puzzle should stop you from doing most things that don’t contribute to the ultimate solution of the puzzle. It shouldn’t block you from doing all things, since then it would be a series of forced moves instead of a puzzle, but it should stop you from going down long dead ends. If there are dead ends, it should be clear that’s what they are when someone gets there. There’s very little that’s more frustrating than trying to explore a huge possibility space where you can’t quite rule out that you’re on the right path.
• Legibility
  For a player to be able to do all this with the puzzle, they need to be able to ‘read’ it. In the Monster’s Expedition example above, you can see that the individual objects are clear, and pains have been taken to make even the grass show where the grid is. If that’s not enough, there’s a setting to turn grid lines on, which makes it all even clearer. It’s very easy to make a puzzle that’s less legible, and while that does increase the difficulty, it’s not in a fun way. As little as possible should get between the player and their view of the possible paths that puzzle can take.

Epiphany Gameplay, Part 1: Epiphany Puzzles

Not all puzzles are about going down a path, solving minor problems as you go. Some puzzles are about having a key insight. These come in a few different varieties, at least that I’ve seen, but I’ll go with the more foundational one first.

A standard ‘epiphany’ puzzle looks impossible at first blush. It violates everything you know about the rules of the game and how the puzzle is supposed to go. Very often these involve a red herring path, which it becomes clear is never going to lead to anything.

The puzzle is unsolvable until you have a key insight into the solution.

For a pure epiphany puzzle, I think this shouldn’t include any new rules unknown to the player, just a reorientation of either their toolset or their understanding of the puzzle space. These sorts of puzzles are distinct from puzzles that introduce new rules or mechanics: they’re puzzles the hinge around a single insight.

Epiphany puzzles are much harder to design than ‘path’ puzzles, because if the player doesn’t have the epiphany, they’re stuck, but if the epiphany is too obvious, there’s no moment of breakthrough and you’re left with a puzzle that basically solves itself. These are the kinds of puzzles that make a player give up on a puzzle game, especially if it’s a mandatory puzzle blocks all other progress.

This is a very early ‘epiphany’ puzzle from Cosmic Express, shown with its solution. The goal of the game is to deliver monsters to their ‘home’, which are indicated by boxes. The train has only two seats. Up until this point, the puzzles have had you picking up monsters and delivering them almost immediately, but the ‘epiphany’ needed to solve this puzzle is that you should pick up only one of the first two monsters, go all the way to the other end of the map, and deliver him to a ‘home’ that’s further away. This is a foundational technique for the rest of the game, and furthers an understanding of what puzzles might require, but introduces no new mechanics. Crucially, it cannot be solved without this insight, and an inexperienced puzzle player might get stuck here trying to do all the obvious things they’ve done before, which is the hallmark of ‘epiphany’ style puzzles.

While a hallmark of epiphany puzzles is the epiphany, a good puzzle designer can give subtle hints and signposts, especially if the other puzzles in the general area have followed good puzzle design. The player might get stuck trying the usual things, but if there’s been conservation of detail thus far, they’ll notice that there’s a weird corner or a wide open space that other puzzles haven’t had before, and that gets them in the right vicinity of the epiphany.

Generally speaking, I think epiphany puzzles should be placed toward the end of a themed set of puzzles, when the player is most likely to have fully learned and internalized whatever the theme is. This makes the epiphanies easier to have, and also means that there are expectations and solving patterns that have been set up to be subverted.

Epiphany Gameplay, Part 2: Epiphany Rules

Under my definitions, an epiphany puzzle doesn’t require you to learn new rules or mechanics. When the puzzle does require you to learn new things, those are epiphany rules.

I’m almost certain that The Witness didn’t wholesale invent the idea of teaching rules through simple puzzles, but it was definitely incredibly influential, and I think you can see the inspiration in all sorts of games that follow it. It’s a very compelling way of learning the rules of a puzzle game, and especially once a player has been taught that this is how rules are learned, I think it offers lots of possibilities for the future.

The simplest epiphany rules aren’t really epiphanies at all, they’re just completely (or almost completely) forced puzzles that directly demonstrate a rule. You are forced to push a blue block onto a blue square, and you get to see that it lights up and the level ends. In my opinion, this is the simplest use of epiphany rules, and is much more effective than simply popping up a screen that says ‘put the blue block on the blue square’, but it’s really not where the approach shines.

There are two places where I think epiphany rules are really stellar, and why they’re one of my favorite things in puzzle games.

Rules You Didn’t Know Were Rules

A properly designed puzzle game can hide interactions from you for quite a while. All they really need is to ensure that a certain interaction never happens, or at least never happens naturally, and then at a certain point, either force it or allow it to happen. I won’t spoil any examples of this, but if you’ve played modern puzzle games, you probably remember at least one of them. You’re solving a puzzle as normal, go to do something you’ve done a hundred times before, then BAM! something unexpected happens. You’re left reeling, thinking to yourself ‘was that always possible’ and ‘what does this mean’ and ‘does this open up possibilities in previous puzzles’. It’s a rush, and most games can only pull it off once or twice: speaking practically, these are moments that the game needs to be designed around.

I’m a huge fan of ‘moment’ game design in general, so it’s probably not a huge surprise that I would love these. Unfortunately, they can only happen once for each player, and depending on the design, they can be ‘spoiled’ by sequence breaking. A lot of my favorite puzzle games just can never be done justice on replay, because the moments that I loved most about them won’t land the same. I find myself watching Let’s Plays to get a vicarious thrill instead, which is a poor substitute.

Unfolding Rules

Sometimes a puzzle game will tutorialize a new element, but I think this is best done with some care put into not revealing everything. You come across a panel or puzzle or island or whatever, and it has something you’ve never seen before, some symbol or object that clearly does something.

Then, you get to figure it out, typically slowly, over the course of many puzzles. Any good rule will have wrinkles and corner cases, different ways that it interacts with other rules, things to explore.

I love that exploration, and figuring out a new rule is generally my favorite part of any puzzle game that doesn’t just flatly explain everything to me. In narrative terms, if the average puzzle is a whodunit, one of these sort of levels is a howdunit. You get the solution, but you have to figure out why it was the solution, which turns the who approach of puzzles on its head.

Sometimes, this approach can backfire. When I was playing Bonfire Peaks I think I ended up doing a few levels out of order, and got to a stuck on one particular level that required me to use a mechanic that I hadn’t even known existed. At least, I assume that I did the levels out of order, since it wasn’t entirely clear to me. I did eventually figure the level out, but it was a slog and left me feeling pretty unhappy.

Perhaps even more than rules you didn’t know were rules, this sort of puzzle has zero replayability, and often less than zero replayability given that once you know the rules, it feels like filler. I’ve played through The Witness four times now, and all three times found the tutorial puzzles to be a bit of a chore. There’s no good solution here, since I’ve never seen a puzzle game with a NG+, and don’t even really know what that would look like.

Themes and Worlds

There’s a pretty classic puzzle game trope of having “worlds” and “themes” and “biomes”. Different areas explore different mechanics, whether those are specific in-game objects with different rules, or specific rules interactions or corner cases.

I think this is a great and sensible thing, and expect that any puzzle game of a certain level of complexity would be better off adopting it. Dividing puzzles into worlds has a few key benefits:

• It allows players to build up knowledge and mastery, and working within a certain theme is probably how they’d prefer to approach puzzles anyway.
• It allows players to know what sorts of puzzles are there, if they feel like doing one thing instead of another (such as if they need a break from blue triangle puzzles).
• It helps foster iteration and building of complexity, partly because of that built up knowledge and mastery.
• It helps create good ‘gates’ to future content, since the puzzle maker can ensure that the player has sufficiently been tutorialized on certain concepts (though this can also be done by having ‘locks’ that can only be solved through having the right knowledge).

Once you have a puzzle game separated out into different worlds with their own themes, the next obvious step is to mix those themes a bit, bleeding red world into blue world, or maybe making that a world of its own. For most players, getting an overlap of themes has a lot of conceptual power, and I think most puzzle games will at least try to have a ‘rainbow’ world that incorporates every mechanic in the game, solvable only with mastery of all the elements.

The drawbacks to this approach are mostly the same as the benefits: you’re signposting what the levels will be like and how they’ll be solved. This can dull some excitement and make the puzzles easier or rote in a way that isn’t always the greatest, especially for certain obligatory puzzle game elements like tutorial puzzles. Generally, I think the drawbacks are worth it, but I think it should be noted that a “world” or “theme” approach does come with design constraints.

Mastery and Player Progression

For my favorite puzzle games, I feel like I’ve grown as a person by the time the game is finished. This is a false feeling, I’m well aware, but a good puzzle game gives you things to learn and master. Generally speaking, these come in two varieties: rules and techniques.

Rules mastery is easy enough: do you know the rules? Rules tend to get learned slowly over time, and even if you learn a rule, you often have to learn how it interacts with other rules, but at the end of the game, and usually well before the end of the game, you know everything there is to know about how the game works. In theory, anyway. There are some games, usually ones that aren’t favorites, where I walk about thinking “you know, I still don’t know if I know how this rule actually works”. Sometimes this happens because there’s some ambiguity remaining when all the puzzles are done, and sometimes it’s just an obtuse rule you can kind of feel your way around like the proverbial blind man with an elephant.

Technique mastery is different. It’s not about the rules, it’s about the implications of the rules. Technique is how you solve the puzzles. At the start of this post, I went into what I consider the basic technique: walking down different paths and seeing where they take you, looking at the possibility space and pruning it down. But there are other techniques that help with the solve, and they’re usually dependent on the individual puzzle games. Some examples of what I mean follow.

This is Q Remastered, a game where you draw on the screen to make simple objects, which are supposed to accomplish a wide variety of objectives. In this case, you can probably guess which of those objects I drew: it’s a hook with a ball at one end. Once I release the mouse, the weight will drop, and the ‘hook’ will lift up the box, completing the level. Q Remastered has a few of these, most of which will occur to anyone playing for more than a few minutes, but it’s a puzzler’s instinct to name and conceptualize these for reuse later.

There isn’t any specific weighted hook rule. The technique is instead an emergent property of the rules. Lots of games have something like this in them, especially if they have simple rules and a mathematical basis. Sudoku is a classic, and has all kinds of interesting techniques, even though there’s really only a single rule of sudoku (one each of the set one through nine in every column, row, and box). Technique comes from things like x-wings and the Phistomefel Ring and Set Equivalence Theory, and once they get added to the toolbox, they’re there for good.

To give a more recent example, and from something that’s not a traditional puzzle game, Tears of the Kingdom has two mechanics, one where you can move objects and another where you can reverse time. Because you can move objects vertically, this means you can move an object up, let it fall, get on it, and reverse time in order to gain pretty significant height. Again, the elevator technique isn’t a rule, it’s a consequence of the rules, but it’s such a major consequence and so reusable that it arguably has as much impact as a rule itself, and can be treated as one for most intents and purposes.

A good puzzle designer can design puzzles that teach techniques to players, and then use those taught techniques in future puzzles as part of a theme or to add some complexity. However, unlike with pure rules, it’s difficult to stop players from using techniques and eliminate them from possibility space.

Corner Cases

Before we get to the end, there’s one thing that I want to harp on a bit, because it’s something that I see a lot, even in relatively polished games built by people who clearly know their puzzles.

Puzzle games are programmatic. They have to be, because they’re programmed. Until you play a puzzle game with a human judge or some weird LLM system, the rules will always be concrete and unambiguous … but this doesn’t mean that’s how they present to the player.

Sometimes, in the course of programming a game, the programmer has to make an arbitrary choice for how something will resolve. Maybe two effects would logically happen at the same time, but can’t actually do that. If there’s a red rule and a blue rule, the game can’t always support having a purple rule to cover that. So instead, the programmer has to decide if red or blue will be first, or if there will be some outside reasoning for which gets their turn.

It is my contention that while this is an unfortunate reality of programming games, every effort should be taken to minimize the presence of arbitrary resolutions, and those resolutions should never be part of a puzzle solution. It’s my belief that most puzzle games benefit from exploring the far reaches of their ruleset, exhausting every implication and interaction … but not like this.

The problem is twofold. The first issue is that most puzzle games present things graphically (putting the video in videogames) and most compromise corner case stuff has no clear visual representation. This means you need to know and remember the rule, which increases cognitive load and doesn’t generally feel nice, at least to me. The second problem is that these corner case rules tend to be inelegant. They’re there because they need to resolve an inherent issue with the other rules, not because they’re fun, clean, pretty, or cool.

I think when I’ve seen them included, it’s out of a desire to get the most possible mileage from the puzzles, to turn over every rock and explore every corner. I think that’s generally a good instinct … but not here, not when it goes against so many other principles of good design.

Puzzle Game Recommendations

In no particular order, except for my favorite game first:

• The Witness
• Baba is You
• A Monster’s Expedition
• Snakebird
• Snakebird Primer
• Taiji
• Lingo
• Can of Wormholes
• Handshakes
• The Talos Principle
• Filament
• Patrick’s Parabox
• Toki Tori
• Linelith

And some games that don’t get a recommendation per se but which I found to be fun enough, and definitely draw from the same conceptual pool, and might be up your alley:

• Bonfire Peaks
• Cosmic Express
• A Good Snowman is Hard to Build
• Sokobond
• Railbound
• Puddle Knights
• Bean and Nothingness

And some other games that I wouldn’t consider puzzle games in the same way as the ones I’ve been talking about, but which do share at least some notable elements:

• Outer Wilds
• Fez
• Braid

There are lots of other games which I would guess are technically puzzle games, just not in the way that this blog post was talking about. In particular, there are ‘programming games’, which I generally enjoy but don’t really fit this mold, like all the Zachtronics games. There are also automation games, which always have emergent puzzles, but which aren’t in the same realm. If my favorite game is The Witness, then Factorio is my second favorite, but I would raise an eyebrow at anyone whose categorization scheme puts Factorio in with the puzzlers.

Conclusion

I personally think we’re in the golden era of puzzle games, and it’s my hope that it continues. Puzzle games are tricky beasts, and puzzle design is difficult and time consuming, at least if you want to get it right. They have the advantage that they don’t need terribly complex programming or amazing art, but that means the main gameplay needs that much more care and attention.

I do worry about how evergreen the genre is. I’ve definitely played enough puzzle games now that I can see certain patterns a bit early, ‘Ah, that’s parity’ or ‘This is tetrominoes’ or ‘Classic sokobon problem’. There’s a lot of space to work with, but some of the simplest and more elegant concepts feel like they might become so well worn that they won’t do anything for me.

If you know of any puzzle games that seem like they might be up my alley and weren’t mentioned here, let me know in the comments. It’s a pretty niche genre, but the fairly low barriers to entry mean that there might easily be hidden gems I’ve just never heard of.

And lastly, a confession: I’ve never played Stephen’s Sausage Roll. I know, I know. It’s on my list.