Embedded Puzzle Manifesto

You might have seen this already on Planet-IF, but here’s the full text of my manifesto on puzzle design (what puzzles are and what makes a good puzzle and how puzzles are most commonly bad). It applies to all game types, but interactive fiction most of all.


A puzzle is a problem with a nonobvious solution.

A satisfying puzzle is multistep: the solution to the problem requires solving a different intermediate problem.

Creating puzzles is easy: your protagonist wants something, problematise them getting it. And then problematise the problem. Repeat until sufficiently complex.

A puzzle is nonobvious. Using a key on a door is not a puzzle. Finding a screwdriver to unscrew the hinges might be a puzzle. Using the butter knife as a screwdriver to unscrew the hinges is a puzzle.

There is no necessary distinction between a narrative game and a puzzle game: puzzles should reveal the narrative.

Puzzles can be dramatic. A puzzle and a dramatic situation are both impediments to the protagonist’s goals. The solution to a puzzle, like the resolution of a drama, tells us something about the protagonist’s character. “Really good at sliding-tiles” isn’t an interesting character facet. “Prepared to steal and lie” might be an interesting character facet. “Will destroy gifts made by loved ones in order to finish this thesis” is an interesting character facet.

Moreover, a puzzle can serve the role of gating narrative until the player understands the dramatic situation. For instance, if they need to know the name of a suspect in order to look them up in the phone book. Or they need to have visited a place or spoken with a character as a part of a puzzle-chain, and those places or people, or the items they had to pick up or disturb, are later narratively important.

It is clear how to make puzzles conflict with a narrative. Here’s how:

  1. Encourage the player to perform unmotivated actions because they know its a puzzle and puzzles must be solved.
  2. Include elements at odds with the setting because of a need to include puzzles.
  3. Include solutions that require out of game knowledge.
  4. Rob the story of urgency by encouraging the player to wander between multiple locations repeatedly in search of a solution.

Thus, it is clear how to make puzzles cohere with a narrative. Here’s how:

  1. Bound the actions available to the protagonist’s motivations.
  2. Embed puzzles in the setting of the game.
  3. Contain solutions in the setting of the game.
  4. Ensure that the intended player experience is the one that you are enabling.

To unpack:

1 – Motivation

The protagonist’s actions should be motivated by their goals in-the-fiction, and as such what the player can make them do should cohere to these motivations. In some games fiddling with everything is motivated by a quest for greater understanding. But in most games, the character has specific desires and it makes no narrative sense for them to solve nearby problems just because they look like puzzles. In real life and in stories, we don’t try to break into the liquor cabinet until we want a drink.

2 – Embeddedness

Unembeddedness is rampant in games with puzzles. There are anachronistic elements introduced in a spirit of zaniness. There are machine systems or locking mechanisms that closely resemble classic puzzles like pipe-turning games or the Tower of Hanoi. These aspects are artifacts of arbitrariness. They take the player out of the fiction, and force them to experience the game as a puzzle set by the game creator.

Embedding a puzzle is about making the difficulty that faces the player believable in the setting while not boring or trivial. A rope across a door is an unbelievable impediment; a locked door is a boring impediment; a snowed-in door that opens to a wall of snow is a believable and interesting impediment.

3 – Containedness

The player shouldn’t need to know the rules to baseball to proceed in your fantasy cavern crawl, or know specific English idioms in your futuristic platformer. If they do need to know these things, they should be contained within the setting of the game. If you need to search a haystack to find a needle you didn’t already know was there, then a character somewhere in the game needs to use the phrase “like finding a needle in a haystack”. And even then, this would only work in a fairytale setting where objects were expected to be placed by idiomatic logic.

What might seem obvious to you as a writer may not be obvious to someone from a different culture or educational background. Moreover, it better serves a narrative if the player is using knowledge their protagonist is supposed to have, rather than knowledge the player must have. Puzzles shine when they act also as a mechanism for ensuring that the player knows what the protagonist is meant to know at that point in the narrative.

4 – Experience

Unless your intended game experience is “frustrated wandering”, implement hints or dynamic events such that the stuck player is recognised by the game and guided before they’re encouraged by the game design to try everything with everything everywhere. An expected flow for an adventure game might look something like this: the protagonist has a problem; the player recognises a number of sub-goals that must be achieved in order to solve this problem; the player directs the protagonist towards these sub-goals; they work on one set of problems whenever they’re temporarily stumped on another set and in doing so usually find a hint that helps them elsewhere. In this expected flow, the problems are nonobvious (i.e. not immediately solved) but the player is rarely bottlenecked and always has something that they can be working on while even they’re stuck elsewhere. The desired flow will look different for a platformer or a JRPG (random menu combats can be dynamic puzzles when well-formed).

Often we see developers decry puzzles. They don’t want players to leave their game and look up a walkthrough. Or they make puzzles so easy (but often narratively arbitrary) that the player is going through the motions to pad out the game. But there is no need to leave puzzles out of our toolbox. We should not be afraid to create difficult puzzles if they are fair and can be broken down into progressively less difficult steps. When implemented well, puzzles are satisfying, further the narrative, tell us something about the characters and are fun to solve. Let us embrace embedded puzzles!

1 Like

One quibble (maybe it’s two):

A satisfying puzzle doesn’t need to be multistep in a single playthrough – a short and repeatable game where the solution to the puzzle comes from failing, and failing, and failing, and failing, and each time learning can be satisfying.

A satisfying puzzle doesn’t need to rely on in-game knowledge in the sense of knowledge communicated on a single playthrough – a short and repeatable game where the solution to the puzzle comes from failing, and failing, and failing, and failing, and each time learning a little more about the world can be satisfying. Such a puzzle does not conflict with the narrative because the narrative in such a case is constructed out of repeated playthroughs, each offering a “view of the cathedral”.

I’m also not sure about this

If you mean that it shouldn’t just be that a needle happens to be in a haystack, then OK. But I can envisage a game where the fact that the needle was hidden in a haystack would be “adequately motivated” (suppose we are told that an NPC has hidden it “somewhere obvious”). I think we are entitled to assume knowledge of English idioms, to some degree. Or we can’t write.

If a whole game is based around that, then it’s fine. A single riddle can be satisfying. A lot of people get a kick out of replaying the same two minutes of Dark Souls until they get it right. One-move parser games can be great. In a sense, there are still multiple steps there insofar as each failed attempt builds up your mental map of avenues (and deadends) of exploration of possibility.

So of course, you and I would instantly know to look for the needle in a haystack. Indeed, when I playtested the rerelease of Fortress of Fear, I did exactly that. But, say, a Spanish player who was fluent enough in English to read and play parser fiction would find it as arbitrary as the monkey wrench puzzle in Monkey Island 2 (in which, if you haven’t played, Guybrush must use a live monkey as a wrench). If the game-story isn’t about idioms, there’s no reason to gate narrative by the player’s out-of-world knowledge of them, especially when they’re so easy to include in the game (in the room text, in player dialogue etc.)

Of course it’s fine to have games require extensive knowledge of idioms if that’s what a large part of the game is about. In the adventure game Toonstruck the player also has to find a needle in a haystack, but the game is about repairing a machine that works entirely on idiomatic pairings (cloak and dagger, nuts and bolts etc.):

Agree about the “one move” thing. Still not so sure about the haystack!

It seems to me that we have to assume some knowledge about something. We assume, for instance, that our Spanish player understands English. Why not assume knowledge of English idioms? After all, if we hid the needle in a gramophone, we would be assuming knowledge (what a gramophone has, that it has a needle, etc) which, today, not everyone would have. That said, I agree that this will work better if it is consistent, in the sense that the game makes a consistent demand on the player. I wouldn’t personally say that a game that requires knowledge of (say) Latin is a bad game: it could be a good game with a limited audience. That’s OK so long as it’s clear that I’m not going to get far if I don’t know any Latin. What becomes unfair is if a game which has been mercifully Latin-free for ages suddenly presents me with a puzzle whose solution is obvious … but only to a classicist.

Agreed, different games can make different demands of us, in much the same way as novels do-- though novels don’t prevent you from proceeding to the next chapter until you’ve shown you’ve understood every paragraph.

This section of the essay was partially inspired by a conversation I had with someone who got stuck on the fantasy platformer game Braid as a puzzle required knowledge of The quick brown fox jumps over the lazy dog. It was visually well clued (there was a continuous animation of a fox jumping over a dog), but never spelled out. Hitherto the game hadn’t required any knowledge of pangrams told to English children, and indeed as far as I can tell it was fairly incongruous with the setting. As such, the person I was speaking to who wasn’t a native speaker and wasn’t playing a text game (which one can understand to be about language) got stuck on the puzzle.

That makes sense to me. IMO it’s not so much the demand as the consistency of it. Counterfeit Monkey demands fairly rich knowledge of English vocabulary, which is fine. But if one suddenly presented a puzzle which required such knowledge in the middle of a game that mostly didn’t it would be unfair. So I suppose I’d say “consistentness” rather than “containedness”, but maybe there isn’t much of a distinction.

I think on reflection there’s another element of “consistency”: puzzles should not be arbitrarily inconsistent. So for instance if puzzle A requires a particular action which would also be feasible for puzzle B, it’s no good if puzzle B doesn’t “recognise” the action. Some of the best puzzle games (Counterfeit Monkey, Hadean Lands) create a consistent puzzle mechanic which enables players to learn an approach, gradually becoming ever more complex. But even if that’s not the objective (and there’s surely room for other approaches), some measure of consistency of approach is desirable.

Two further points which relate to the last topic on your manifesto:

  • The very worst sort of “wandering around” involves active misdirection (a typical neophyte example involves a puzzle which /is/ solved by something approaching violence, but where the standard parser response to HIT has not been caught).
  • Quite apart from explicit hinting, good puzzles recognise honest attempts and provide responses which are not mere refusals, but provide additional information which will lead towards a solution. That doesn’t have to be a direct pointer to what will work: it often just has to be a decent explanation of why the thing you tried didn’t.

I think someone pointed out (is it Graham Nelson in DM4) that since the experience of puzzles is /mostly/ of failure, it’s important that the failure be interesting too. “Ha ha that didn’t work” is merely frustrating, unless it prompts either amusement in itself or some further line of approach.

I was seriously confused until I realized that you meant Fez, not Braid. :slight_smile:

That alphabet “side quest” in Fez is roughly in the tradition of the puzzle extravaganza, which makes much of the idea of applying your background knowledge in surprising, out-of-context ways. “Inconsistency” is a core element of the genre! See also those casual puzzle games where you have to think of every UI element on the device, including the tilt sensor, the home button, the camera, the game’s “credits” button…

Naturally it’s a tricky balance. It’s almost expected that people will work in groups (puzzle hunt teams or game forum groups) so that everyone’s varying backgrounds will fill in each others’ gaps. It’s also expected that, sometimes, you’re going to walk away from a puzzle saying “Yeah, that one was bullshit.”

Your post was really interesting, and made me focus on something I don’t think I’d ever consciously realized before – having to do with your idea of “not breaking into the liquor cabinet til you need a drink.”

I realized I tend to fall into a sort of uber-goal/sub-task mindset when playing games, wherein the uber-goal is clear and makes sense, and the sub-tasks themselves may be meaningless, but I know I’m in a game and these sub tasks must be important for the uber-goal. For example, if the uber-goal is “escape from this house” or “return the gem to its altar” or whatever…in a real house, I rarely feel the need to search every cabinet and every desk drawer, especially ones that are locked. But in the context of the uber-goal, I am definitely going to open veery box along the way – including the puzzle box that needs to be solved to be opened – because I suspect it will arm me with what I need to solve the “main” puzzle (or, why would it be there?). In fact, I’m realizing that games that have stuff that can be elaborately interacted with but aren’t in some way necessary to solve the overall game tend to annoy me.

I’m curious whether you would argue that these “sub-tasks” are sufficiently motivating if (and maybe only if) the case is clearly made that they are steps toward your main goal?

Or rather are you advocating that each and every puzzle in a game should have it own, clear story reason to interact with it at the time you encounter it, separate from (though potentially related or subservient to) the main goal?

Not sure if I have an opinion in which is “right”, just curious your perspective on it.

Thank you for the thoughtful and thought-provoking post!

I’m a big fan of emshort’s analysis of best puzzle. xyzzyawards.org/?p=386

She identifies eight features of puzzles: Extent, Explorability, Surprise, Ingenuity, Originality, Fairness/accessibility, Structural integration, and Narrative integration. But this classification isn’t a manifesto.

I don’t think I especially agree that “satisfying puzzles need to be multistep”, which is to say, you don’t need to have Extent or Explorability if the puzzle has a lot of Surprise (which still being Fair).

Yeah that’s spot on. Anything that increases the faith in the player that their time isn’t being wasted.

Ah good catch. One of the reasons I treid to keep it broad brush without invoking specific examples is this tendency to misremember. Doesn’t help I haven’t played either.

Definitely there’s room in this world for those kinds of games. Having looked into a bit more, the Fez example wasn’t apropos to what I was talking about as an Easter-egg hunt is very different to a narrative game.

It’s relatively easy to give a narrative reason for exploring every locked drawer, looking under every rug etc. if that’s what you want your game to be about. But what runs counter to the narrative is when the player is encouraged by the game’s structure to do that stuff even when it makes no sense for their character in the story. Making these substeps relevant just takes a line of prose or dialogue (indeed, a lot of adventure games make a stab at this with the common line “this might come in handy later” whenever the character picks up some item in the environment when there’s no need for it).

Sure, I’d go along with this. The typical satisfying puzzle is a part of a puzzle-chain, but it’s not necessary or sufficient.