What makes a game worth playing?

Begin with a moment, not a feature list

Imagine a tiny courier carrying the last letter of the evening. Home is across a flooded garden. A bridge is quick but takes lantern oil. A stepping-stone route is slower, and the water is rising. You have enough light for one mistake. Which way do you go?

Now imagine a different pitch: an inventory system, procedural weather, six biomes, customizable outfits, crafting, and a day-night cycle. It might describe the same game. But it tells you almost nothing about what a player will actually enjoy. A feature list describes the machinery. The first scene describes a situation someone can care about.

We will call our little courier game Night Mail. It is an original teaching example, not a product you need to buy or a project you need to finish. Throughout the course, we will change its rules, test its choices, and occasionally discover that our clever idea is a nuisance. That last part is normal.

Start your own notebook with a moment. What is happening? What can the player do? What are they trying to understand or accomplish? If your idea is “a cozy farming game,” zoom in. A baker has two ripe crops and one free oven slot. A neighbor arrives before sunset. Now you have a situation to work with.

Try to learn the signal

The courier reaches a signal post. Your job is to predict its next signal. Try each mode for several turns. Pay attention to your own thoughts, not just the score. When do you form a theory? When do you stop needing one? When does another attempt feel pointless?

These are deliberately tiny experiments. They do not measure whether you are good at games, and a correct answer is not a measure of fun. The point is to make three different relationships between action and information easy to feel.

What changed between the three versions?

In the repeating version, learning finishes almost immediately. You can keep answering, but the next answer asks little of you. This can be pleasant for a moment, like tapping a familiar rhythm. If the whole game depends on discovery, though, there is nowhere left to go. A counter that keeps climbing does not create a new question by itself.

In the rule-based version, a wrong prediction can still be useful. You notice a relationship, try it, and refine your explanation. A successful guess feels different when you can say why it worked. You begin to own a small piece of the system.

In the opaque version, there is a sequence, but the relevant rule is not available to you. Another attempt gives you a result without giving you a reliable way to improve the next decision. This is an important distinction: a system can be completely deterministic inside the code and still feel arbitrary to a player.

Raph Koster’s A Theory of Fun connects enjoyment with learning and recognizing patterns. It is a useful lens here: ask what the player is getting better at understanding. It is not a universal checklist, and it does not mean every enjoyable game must behave like a puzzle. The experiments and examples in this course are our own applications of that question.

Write down a sentence about each mode. Avoid “good” and “bad” for now. Try “I knew the answer but kept pressing to see the counter,” “I changed my theory after the third signal,” or “I could not explain why I was wrong.” Specific descriptions give you something to design with.

Learning is bigger than solving a puzzle

In a platformer, you learn how long a jump travels, how the camera follows, which surfaces hold you, and how to recover when a landing goes wrong. Some of that knowledge becomes physical. You stop calculating the jump in words, but your fingers are still making a prediction.

In a farming game, you may learn a daily routine: visit the plots before the mill, save one ingredient for the neighbor, leave room in the afternoon for exploration. In a social game, the thing you learn may be another person’s habits. In a building toy, it may be what shapes can support an idea you have not quite found yet.

A player can also enjoy a familiar action because it is expressive, soothing, social, beautiful, or emotionally important. Someone rearranging a garden does not necessarily want the garden to become harder. Someone revisiting a favorite place may want recognition rather than novelty. Do not “fix” that experience by adding a test they never asked for.

For Night Mail, we might choose quiet competence as our central experience: “I am becoming the person who knows this garden after dark.” That gives us useful direction. Landmarks, readable routes, and opportunities to use remembered shortcuts help. A random teleport that changes every path each night might undermine the experience, even if it adds variety.

This is why you need an experience in mind before you judge a mechanic. The same uncertainty that makes a dangerous expedition exciting can make a relaxing planning game irritating. The question is not whether randomness is fun. It is what this uncertainty asks this player to do, in this moment.

Look for a prediction, an action, and an answer

A useful way to inspect a small moment of play is to follow the player’s thinking. First they notice a situation. Then they form some expectation, choose an action, and read the result. That result changes what they believe or what they can do next. You do not need to display these steps in the interface. You need to make them possible.

At a Night Mail junction, the courier sees water lapping against the lower stones. The player predicts that waiting will make the route worse. They take the bridge, spend oil, and reach dry ground. The next junction arrives with less light available. The action has both an immediate answer and a consequence for a later choice.

Compare that with a junction whose two paths have identical travel time, identical cost, and identical views. The player can click left or right, but the choice does not help them express a preference or learn anything relevant. You might keep it for wandering or visual variety. You should not count it as a deep routing decision.

Feedback is the answer the game gives. It can be movement, a changed resource, a sound, a character reaction, a newly visible route, or an altered relationship. A burst of particles can draw attention to that answer. It cannot replace the answer. If a package vanishes, the player still needs to know whether they delivered it, dropped it, or hit a bug.

Draw four boxes in your notebook: situation, expectation, action, result. Fill them in for ten seconds of your own game idea. If one box is vague, resist adding features. Make that moment more legible first. A small loop you can explain is more useful than a large loop you can only advertise.

Play it: find a rhythm you want to keep

A moving cup. A falling ball. One input. This is enough to ask a surprisingly useful question: when does a repeating action stay interesting? Open the game and try to land three balls. You are learning how far the cup travels while the ball falls.

Start on Slow. Once you can explain your timing, switch to Quick without changing your plan. Adjust after each miss. Finally try Wild. Keep track of whether you are anticipating a result or pressing and hoping. The label is not a difficulty recommendation; different players will find a satisfying pace in different places.

• Predict: where should the cup be when you release, rather than when the ball lands?
• Compare: after three attempts at each pace, which version gives you a useful lesson when you miss?
• Make it yours: name one change that could add a fresh decision without simply making the cup faster. A second landing zone? A limited number of drops? Explain what the player would consider.

A new costume is not always a new challenge

Suppose Night Mail has a puddle you can cross with a short hop. The next level has a purple puddle. The third has a sparkling purple puddle. If the player solves all three by holding the same button for the same duration, you have added visual variation. That can be worthwhile, but you have not asked the player to rethink the action.

Now place a low branch over the third puddle. A long jump hits the branch; a short jump clears the water. One familiar action needs a new judgment. Later, place the branch above a moving raft. The player has to combine timing with the earlier lesson. The new situation draws on something they understand and makes that knowledge useful in another way.

Depth often comes from relationships like this. You can add it by changing the arrangement of existing pieces, the information available, the cost of a mistake, or the goal. You do not always need another button, resource, or enemy type. Every new system also becomes something the player must learn and something you must maintain.

There is a limit. Combining every obstacle at once can turn recognition into noise. A branch, moving raft, collapsing bank, hidden wind, and disappearing destination may create five sources of failure without helping the player separate them. The question “what should I try differently?” becomes hard to answer.

Take one obstacle from your idea and invent three variations. In the first, change only its appearance. In the second, change one decision the player makes. In the third, combine it with something already learned. Explain what knowledge transfers between versions. If your explanation is only “this one is harder,” keep looking.

Separate a worthwhile challenge from friction

A difficult route may ask you to plan carefully and execute a precise jump. A frustrating route may hide the landing, accept inputs inconsistently, and send you through a long menu after failure. Both can produce a low success rate. Only one necessarily supports the experience you wanted.

This is why “players died five times” is not enough evidence. Watch what happens between attempts. Did they change the jump timing? Did they notice a clue? Did they understand the mistake? Or did they repeat the same input because the game never explained what failed?

Accessibility belongs in this distinction. A player may understand the route perfectly and still need a larger target, a slower input window, remappable controls, or a cue that does not depend on color. Offering those options can preserve the interesting decision while removing an unrelated barrier. Do not treat every source of difficulty as part of the game’s identity.

For our courier, deciding whether to spend oil can be the intended challenge. Reading tiny gray text that tells you how much oil remains probably is not. Fixing the text does not make the decision shallow. It lets more players reach the decision in the first place.

What if the player does not want another challenge?

A person can enjoy arranging flowers without needing to beat a flower-arranging boss. They might want a place that feels like theirs, a task that settles their attention, or a small gift to make for a friend. If your only design question is “how do I make this harder?”, those pleasures become difficult to see.

Przybylski, Rigby, and Ryan’s motivational model of video-game engagement draws attention to competence, autonomy, and relatedness: feeling capable, acting with a sense of choice, and feeling connected. Treat these as useful questions about an experience, not three meters that guarantee enjoyment when filled.

For Night Mail, imagine three optional activities after a delivery. A practice route lets the courier improve a jump. A garden corner lets them arrange plants with no score. A noticeboard lets them leave a route suggestion for a friend. These activities offer different reasons to care. Adding all three would increase the project size; identifying the reason you actually want to support helps you choose one.

Consider the garden corner more closely. If every arrangement receives the same noisy “PERFECT!” badge, the response may flatten the pleasure of choosing. A quieter response might let the player see their composition in evening light. If the goal is expression, recognition of the particular result can matter more than a generic reward. The sunflower has survived without a leaderboard so far.

Now remove the daily reward from your proposed activity. Would anything remain that you would want to do? Perhaps the movement feels good, the arrangement is yours, or a friend will see it. If nothing remains, that is a useful design finding. A reward can mark progress, but it cannot tell you which part of the activity is worth preserving.

Write two versions of your experience promise. One should describe becoming more capable. The other should describe a pleasure that does not require improvement. You do not have to serve both. Choose deliberately, and ask a playtester about a specific moment that might support your choice.

Look at the same moment through a different window

Jesse Schell’s The Art of Game Design uses lenses to encourage designers to examine a game from different perspectives. The useful habit is changing the question you ask. You do not need to memorize a deck before making your first game.

Here is our own three-pass exercise. First describe a Night Mail crossing as a player: “I am carrying the last letter and I cannot tell whether I have enough oil.” Next describe it as someone watching: “The player opens and closes the map three times.” Finally describe it as a maker: “The remaining oil appears on the map, but the bridge cost only appears after closing it.” Each view catches something the others can miss.

Write the three accounts before proposing a fix. Otherwise you may interpret every observation as support for the first idea that occurred to you. In this case, putting both costs in one view is a testable repair. Making the bridge cheaper would answer a different question.

Keep this exercise small. One crossing, three accounts, one possible change. A useful theory earns its place by helping you notice something you would otherwise miss.

Write a promise you can test

“My game will be fun” is impossible to use during a difficult design decision. “Players will feel clever when they save oil by recognizing a safe shortcut” is more useful. It names a feeling, an action, and a situation you can put in front of someone.

Keep the promise small enough to test before making a whole game. Build two routes and one visible oil counter. Give the player a reason to care about the remaining oil. Watch whether they identify and use the shortcut. You have a test of one idea, not proof that the entire game will succeed.

Be willing to discover a different appeal. A tester might ignore efficiency and spend ten minutes arranging lanterns along the path. That does not automatically mean you should build a decoration game. It does mean you have observed a behavior worth asking about. Your promise is a direction for inquiry, not a contract that the player must obey.

1. Write one moment of play in three sentences. Include what the player sees, does, and changes.
2. Name the experience you hope that moment creates: curiosity, mastery, relief, expression, connection, or something more specific.
3. Write what the player might learn or express through the action.
4. Choose one visible behavior that would suggest your idea is working. Include a behavior that would make you doubt it.
5. Cut the experiment down until you can try it with paper, a few shapes, or an existing editable scene.

A worked answer, including what could go wrong

Promise: “I want the player to feel resourceful by using a remembered dry route when the garden floods.” Small test: one map, two deliveries, one route that becomes unavailable, and landmarks that stay in place. Evidence to look for: on the second delivery, the player notices the flood and tries a remembered alternative without needing a route arrow.

Possible failure: the player remembers the route but cannot tell that the lower path is flooded. That points toward clearer state communication. Another failure: the player understands the flood but finds the alternate walk tedious. That points toward pacing or a more interesting route. Those are different problems. “Add more fun” would not tell you which one to solve.

Another possible result: the player feels anxious about losing the letter, while you intended a gentle evening walk. You could reduce the consequence of falling, provide a nearby recovery point, or remove the deadline. You could also decide that mild tension belongs in the game. The important part is making that decision deliberately.

You now have the beginning of a design practice: describe an experience, build a situation that might create it, observe what happens, and revise your explanation. The next lesson looks more closely at the choices inside that situation. A game can offer plenty of buttons and still give a player very little to decide.

Sources & further reading

Sources for the ideas discussed here, plus further reading. The lesson’s examples and interactive exercises are original.

• Raph Koster: A Theory of Fun for Game Design
• Raph Koster: An atomic theory of fun game design (historical essay)