Making puzzles hard on you

How to make difficulty fun?

For the last few months, I’ve been working on my second PICO-8 game: Dank Tomb. The core idea for that game came from a real-time lighting demo, of all things. Since I was starting from the visuals, I had to search for a while for the right set of mechanics to match them. I considered horror themes and a few other approaches before settling on a platforming/puzzle mix with a top-down perspective, similar to how older Zelda games did their dungeons.

Puzzle games live and die on their learning curves and striking the subtle balance between being too easy and being too hard. Stray too far in one direction and you end up losing players to either boredom or frustration.

Knowing up-front that I’m working on a very limited system, I realized that Dank Tomb was going to be small content-wise. I had to milk each room and each puzzle for all it was worth and couldn’t afford making too many missteps in either direction on the difficulty curve.

This made me think a lot about the nature of puzzle games — what exactly makes a puzzle hard, what makes a puzzle fun, and how the two relate to each other. I’d like to share these thoughts with you today.

Making life difficult

Whenever an idea has to be materialized, somebody has to come up with a design. An industrial design for everyday objects, UI design for applications, level design for games.

Game design is a very strange beast — it is, in a way, the exact opposite of other types of design. Ease of use is of tantamount importance when designing a microwave oven or a mobile app. “Challenging” is not really the word you want to hear when people describe your product.

A rather elaborate design for an elevator call button.

In games however, challenge is a crucial component. The game designer deliberately throws obstacles in the players’ way. The players take pleasure in overcoming them. This is especially pronounced in puzzle games — if using an elevator in real life was as hard as it is in Portal, we’d probably all take the stairs instead.

A method to the madness

For a puzzle game to be fun, it has to be difficult. My goal as a designer is to challenge you, but never to defeat you. It’s a bit like romance: I have to confuse you at first, but also subtly guide you towards the correct solution to give you that moment of pure joy when you finally get it.

But difficulty is a complex topic, and a puzzle being difficult doesn’t always mean the same thing. In an effort to better understand what makes puzzles tick, I took inspiration from Mark Rosewater, the lead designer for Magic: the Gathering.

At one point, Magic R&D had trouble controlling how complex their game was, which made players new to the game see it as hard to play. Mark tried to discover the root cause of this problem and ended up identifying three different sources of cognitive load: comprehension complexity, board complexity and strategic complexity. This led to solutions for handling each of them — the details of which you can find in his excellent article.

Mark’s approach inspired me to try a similar thing for puzzle games: to identify what sources of difficulty we introduce when designing puzzle games and the benefits, drawbacks and dangers of using each.

Comprehension difficulty — figuring the game out

This is the first type of difficulty experienced when trying a new game or mechanic. Comprehension difficulty is all about the first question popping up in the player’s mind: “how does this work?”.

Let’s say we have a game based on lasers and mirrors. There is also a sensor-like thing on the wall. It’s pretty obvious what to do here, but we still have to learn how. Can you turn the laser around? Can the mirror be pushed or rotated, or is it fixed in place? Does the color of the beam matter? Does the laser kill you if you cross the beam?

In many modern games, this kind of difficulty is treated as a necessary evil and copious tutorials are introduced to combat it. Sometimes it’s the only way to do it — many mechanics are hard to discover simply by exploring, and there is no other way to teach controls.

But leaning on tutorials has its share of problems. They work by lecturing the player on game mechanics and expecting them to memorize it all. As the word “lecture” implies, this is not only unfun, but also unreliable. People learn more readily if they discover the solution on their own instead of being told what to do.

But we have to teach the players somehow. If we can’t tell them what to do outright, they are forced to simply try things until something works. This turns out not to be a problem, as long as we design the game to accommodate that.

No instructions, but it’s not exactly rocket science.

A good modern example of how to do this is The Witness. You start the game in a confined location with a single puzzle on the door and no explanation on how to solve it. This is not a problem, given a few properties of its design:

• the goal is clear: the only thing in the room is the door
• the location is safe: there is no time pressure, no enemies to worry about and no way to fail
• the puzzle is dead-simple: making it possible to find the correct solution with no information
• the mechanic is intuitive: you are given something to draw with and a labyrinth, drawing a path is the first thing that comes to mind
• it forces you to learn: no way to proceed other than learning the rules

Solving this puzzle teaches you the core mechanic of the game — panels with patterns have to be activated. Upon activating them, you draw a path to an exit. Subsequent puzzles build on this, but the basic idea holds for most of the puzzles in this game.

The Witness is successful in teaching us the rules, even though it doesn’t offer a single word of explanation. To duplicate its success in our own games, we have to generalize and replicate the five things that made it work:

• keep it focused: lock the player in a small location, make the goal clear
• provide safe environment for experimentation: no distractions
• keep it simple: make the minimal puzzle required to show the rules
• make it intuitive: make it easy to guess the behavior of your set pieces
• make it exhaustive: make sure the only way to solve the puzzle is to learn all the rules you expect it to teach

Let’s apply this together to the lasers example. Deflecting light beams with mirrors is pretty intuitive as an idea, but we still have to take care of the other conditions. So we decide to lock the player in a small room (staying focused) with one laser, one mirror (staying simple) and nothing else (safe environment). Our lasers are fixed in place but can be rotated, while mirrors can be both pushed and rotated. The puzzle should require performing all these actions in order to teach the basics properly (exhaustiveness).

In our game, a laser beam also kills the player when they cross it. We could try to include this in the introductory puzzle as well, but that’ll muddle things up and make the original puzzle lose its focus. It’s probably better to design a separate puzzle to teach that and only that.

Conceptual difficulty — beyond the basics

Having shown the basic rules of the game to our players, we can start building on that knowledge. The first of two main ways to do it is introducing conceptual difficulty: forcing the player to use their existing knowledge in new ways.

The player encounters a room with lasers and mirrors, but no sensors. They realize it’s not a standard “guide the beam to the sensor” puzzle. It quickly becomes obvious that to cross the room, the mirrors have to be used to form a clear path through it — otherwise, the beams will kill you.

This is a new concept, if a simple one, that the player has to come up with in order to solve the puzzle. The rules governing each of the components are already clear: lasers kill people, mirrors deflect beams. You only have to connect the facts you already know to get the solution.

That’s the core of conceptual difficulty: making the player come up with new uses for things they already know. This type of difficulty builds on itself. Once the player solves a conceptual puzzle, the concept behind it becomes part of their game knowledge. Using it in future puzzles no longer adds any difficulty.

I’m going to use The Witness again to illustrate the idea of conceptual difficulty, since it’s a game that wears its structure proudly on its sleeve. Puzzle panels in this game usually come in sets. The first panel in each set requires you to figure out a new concept, like realizing that the paths in the puzzles correspond to the trees nearby.

The branches of this tree seem to match the shape of the puzzle. Probably a coincidence.

This type of difficulty can be great for the game. Solving the puzzles leads to a satisfying “a-ha!” moment for the player and maintaining a steady stream of new concepts keeps the gameplay fresh.

But there is a flip side: these puzzles can be dangerously binary. Solving them by definition requires a flash of insight, and it’s easy to design one that becomes three minutes of staring at the screen, followed by finally getting it and then doing it in five seconds. This can be fun in moderation, but it quickly becomes boring if the only thing happening on-screen is the character’s idle animation.

What’s worse, the player can easily become stuck on such a puzzle. If the only way forward for the player when they’re stuck is to stare at the puzzle even longer, there is a big chance this player is already gone and playing another game.

To avoid falling into this trap, this type of puzzle should be designed carefully to allow experimentation — pushing things around, fiddling with knobs, trying stuff. It’s best if such experimentation yields some hints: a partial solution that’s possible to stumble upon randomly, or a non-solution whose failure sheds some light on what needs to be avoided. The geography of the level can also help by highlighting important things, using lighting or a strategically placed ledge to point out the right spot for a mirror.

That way, people who solve it instantly in their head still get to do it and reap all the benefits — they don’t need to experiment and won’t notice your hints. Meanwhile, those who become stuck have a way to get out of their rut and work towards the solution.

Combinatorial difficulty — quantity has its own quality

While conceptual difficulty was about requiring you to come up with a new idea, combinatorial difficulty only uses known concepts, but requires applying them to more things at a time.

We can build a pretty standard room with lasers, mirrors and sensors. This is the standard puzzle design for our laser game — nothing new here. But this time, we’ll put in five of each.

Suddenly, there are new facets to the puzzle. You need to figure out which mirror goes with which beam. A mirror intended for one laser can block another one’s beam, making the positioning trickier. You also have to figure out the proper order — if you pick the wrong beam to start with, it can block off your path to reach another mirror.

This is another idea The Witness spells out for us. As I have mentioned in the previous section, puzzles in this game come in sets. The first puzzle in such a set usually teaches something new. The following ones simply provide more. Larger patterns to fill, longer paths to draw, more things to keep in mind.

If putting the puzzles side-by-side like this “they’re all based on the same rules” obvious, I don’t know what would

This approach is much safer than introducing a new concept. There is essentially no danger of the player getting hopelessly stuck. All elements are known quantities and their workings are clear. It’s just a matter of finding the right sequence of moves.

The trap here lies in relying on this approach too much. Since these puzzles introduce nothing new, solving the fourth one in a row can feel like work instead of fun. The game becomes stale and the player disengages. There is also a limit on how far you can push things — put too many moving pieces in your puzzle, and the player will not be able to keep track of them all.

The solution to both problems is to use combinatorial difficulty in moderation. Grow the puzzles only as long as it stays satisfying, then switch to a new idea, or intersperse puzzles based on different ideas to keep things fresh all the time.

Awareness difficulty — never forget

Combinatorial difficulty was about the size of the puzzle’s possibility space. Awareness difficulty is about a more literal kind of size: the actual dimensions of the puzzle in time and space.

Let’s say you encounter a room with a laser and three holes in the wall. Aiming the laser at them doesn’t seem to do anything, so you let it go after a while and explore further. Ten minutes later, you find a similar room in a different part of the level. This time, there are three holes and a light sensor, but no laser. This forms a meta-puzzle spanning multiple rooms.

When you read about it, it‘s not exactly rocket science: figure out which hole gives access to the sensor, go to the laser room, aim it correctly, return to the sensor room, done. But it’s not as simple in actual gameplay.

In between seeing the two rooms, you probably visited ten others, solved five other puzzles, saw three cutscenes and in all likelihood forgot entirely all about the holes you saw fifteen minutes ago. It might have even happened yesterday, if you just started from a saved state. This is where awareness difficulty stems from — remembering the right things at the right time.

This type of puzzle can be satisfying as it gives a greater sense of place to your levels — big, map-spanning puzzles help make a game world into a cohesive whole instead of a series of disconnected challenges.

But there are also some significant dangers here. First, when this type of puzzle fails, it fails hard. The only way out when you miss a crucial clue is to re-explore places you’ve already been to in an effort to stumble onto something. If you don’t know what you’re looking for, it’s just a random walk until you find something that clicks, the canonical “I’m stuck” experience nobody is particularly fond of.

Second, even if you remember all the facts, there is still the issue of backtracking. If the rooms involved in the solution are far removed, solving the puzzle will take a lot of time — time spent going through rooms you already visited. Some players have built up a tolerance for this, while others will rage-quit at the first hint of being required to backtrack.

There are a few approaches to combat this. You can make the rooms taking part in the meta-puzzle memorable and close together. You can limit the backtracking by introducing shortcuts. You can highlight the “interesting” rooms on the player’s map, or even provide a giant arrow pointing them in the right direction. But these are all band-aids that cover up issues.

But there is another solution, one that works by going in the opposite direction: making the meta-puzzle really big, a theme for an entire section of the game. It’s easier to miss a room with three holes in a wall than a big, shiny machine taking up a whole floor, with various inputs, outputs, pipes and valves sprinkled throughout the level. Such an elaborate puzzle pulls a lot of focus and is instantly memorable.

The franchise most famous for having such puzzles is Legend of Zelda. These games are fond of making entire dungeons into one big puzzle, requiring you to turn the level upside down or smash pillars in order to collapse the ceiling.

Perhaps the best-known Zelda dungeon with this feature is the Water Temple from Ocarina of Time. It features valves that let you control the water level within the entire dungeon. Each setting gives you access to previously unreachable passages while blocking off others. Figuring it all out and making your way through is very rewarding.

The only indication this is not a sewer map is that Link doesn’t usually feature in them. (image from a Boss Keys episode, courtesy of Mark Brown)

However, even the Water Temple — the prime example of how to do this type of meta-puzzle right — is polarizing. Some people love it, others hate the backtracking and the slow pace induced by the sheer scope. Nintendo changed their design philosophy in later games, making dungeons with more shortcuts, less backtracking and simpler meta-puzzles — largely due to this problem, one might think.

Execution difficulty — how practice differs from theory

The final type of difficulty is cut from a different cloth. Execution difficulty is not about how hard the solution is to come up with, it’s about how hard it is to perform in-game.

This type of difficulty is more about dexterity than intellect and comes on a sliding scale that is usually pretty stable throughout the game. The game’s spot on this scale defines the exact genre it belongs to and the target audience it’s intended to reach.

Many pure puzzle games feature no execution difficulty at all. Once you figure out the solution, you can do it at your own pace, no dexterity required. That’s a valid choice, and some players prefer such games, but there are some caveats.

If the puzzle is elaborate, going through the motions of implementing the solution can take a while. Given that executing the solution is trivial, the fun part of the game has already happened in the player’s head. The part happening on the screen is a foregone conclusion and is only required to prove to the game that you solved it. That’s only a step up from a cutscene and can make the game boring in the long run.

The other extreme of this sliding scale is best exemplified by platformers like N, Super Meat Boy or Celeste. Many skill-based platformers are actually hybrid puzzle games in disguise. The puzzle lies in finding the right sequence of runs, jumps, wallclimbs and other moves that is needed to get you to your destination.

Getting there is as hard as figuring out how.

Even once you know the solution, making it happen with the right timing is very hard in these games. If your game’s focus is the puzzles and not the action, that might not be very wise — you might end up frustrating your more cerebral-focused audience.

What strikes me as the most important thing about execution difficulty is that you can choose to have any amount of it — as long as it’s relatively constant throughout the game. If you decide to have no dexterity-based elements at all — fine, there are people who will love that! If you have a lot, there are people who will adore that too.

What’s dangerous is mixing both approaches within one game. The only thing you’ll achieve that way is alienate both groups, with action-loving types bored by the slow puzzles and hardcore puzzle-lovers put off by action sequences.

It’s also important not to mix execution difficulty into hard puzzles of other types. The most common piece of feedback I got from early Dank Tomb playtests was how frustrating it was to solve an elaborate puzzle only to get killed by missing the last jump required.

All together now

So, I’ve got five different avenues for understanding how difficult my puzzle game is. That’s great, but what are they good for?

The main benefit is that naming a concept makes it easier to reason about it. We often feel it when something is wrong with our design, but without a proper framework to think about it, it can be difficult to find the cause. Focusing on the difficulty types helps a lot with some aspects of puzzle design.

That’s how a design for a Dank Tomb room begins.

I can now consciously ask myself — which type of challenge am I trying to pose? Once I figure that out, it helps me stay focused. If my goal is to make a conceptual puzzle, I know that I should probably keep the number of components pretty low. I also double-check how easy the puzzle is to experiment with and if there are enough hints embedded in the design.

Another thing to remember is ramping up the difficulty one type at a time. If I quickly increase conceptual and combinatorial difficulty right after introducing a new concept, I run the risk of turning my game’s learning curve into a brick wall. It’s better to do it gently. First a room that makes the combinatorial space a bit bigger, then another one that focuses on a new idea.

In addition to providing focus to individual puzzles, thinking in these terms helps focus the overall design of the game. For Dank Tomb, I decided to focus on conceptual difficulty, as this seemed most appropriate for creating the feeling of breadth within such a small game. Knowing the caveats this type of difficulty has, I had to order the puzzles carefully to make sure every concept has had a proper introduction before it’s utilized. I made a point of mixing other types of difficulty in. Conceptual difficulty is the main focus, but I intersperse it with rooms based on different principles to engage different parts of the player’s brain.

These categories are also useful during playtesting. If you see a player struggling, you can ask yourself: what is the problem here? Is the player missing a concept? Let’s add hints to the puzzle! Are they having trouble figuring out the right sequencing of actions? Maybe lower the combinatorial complexity. They’re having trouble with a rule because you haven’t taught it properly? Make sure the location that was supposed to teach it does its job and is presented at the right time! Or you know, add one — sometimes you miss things completely because the rules make sense to you.

Wrapping up

As it is with so many things in design, it’s hard to offer iron-clad rules and definite answers when it comes to keeping difficulty fun. All I have to offer are new tools to help us understand what makes game tick and what can be tweaked to make them better.

Your answers on which types of difficulty to prioritize, what their problems are and how to cope with them might all be different than mine. I still hope that whatever they are, my article helps you find them!

If you’d like to see how I put my ideas into practice, Dank Tomb’s beta version is available through my Patreon page — which also helps finance writing more articles. Being a patron is a great way to be notified when I write something new, but you can also follow me on Medium or Twitter!

Until then, may all the games you play be challenging and fun!

This article exists thanks to my Patreon supporters: Tomasz Bylina, Adam Kwapiński and many others.