Lesson 3 · Part 2: The Player

The Other Computer

A game runs on a machine, and also in the player's mind. The mind is the one that matters.

12 min read

The Other Computer

You’re in a hotel room in the dark, holding a TV remote you’ve never seen before. You press the biggest button. Nothing. You press a smaller one and the volume jumps, so volume must live in the top corner. You press the button just below it, expecting the sound to drop, and the channel changes instead. You adjust your theory and press again.

Half a minute later you can work the TV. You never opened the remote or read a manual. What you have is a small private theory of how this remote behaves, pieced together from guesses and corrections, and it’s good enough to predict what the next press will do. Everyone who plays your game builds a theory like that. This lesson is about why that theory, more than your code or your cardboard, is the thing you’re really designing.

A game runs on two computers

Here’s a way of seeing it that a lot of designers find useful.

“As a designer, you’re actually dealing with two computers.”

Will Wright

The first computer is the one the game runs on: a console, a phone or, for a board game, the rulebook and the pieces. The second is the player’s brain, and especially their imagination. That one is far more complicated, and nobody has the manual. The two computers idea is a reminder that your game has to run on both, and that the fun only ever happens on the second.

Take The Sims, a game where you look after a household of little simulated people. To the machine, a Sim is a bundle of numbers: hunger, energy, fun, a mood score. Players don’t see numbers. They see a man who burned his dinner, missed his ride to work and is now sulking on the sofa, and they feel sorry for him. The machine never stored “sulking.” The player brought that.

The two computers also change at very different speeds. Consoles get faster every few years. Brains are the same model they were a thousand years ago, which is oddly comforting: whatever you learn about how people think will still be true when today’s hardware is in a drawer. In Level 2 you met constraints. What a player can take in turns out to be the tightest one you’ll ever work against.

What the machine sees
Walk around the farm and say hello to the animals. Then switch to “Machine” and keep playing. It’s the same game at the same moment, as the other computer sees it.

The picture in the player’s head

A mental model is a person’s working picture of how something behaves, the one they use to guess what will happen next. Your theory of the hotel remote was one. You carry thousands of them: for doors (push or pull?), for traffic, for how your grandmother reacts when you’re late.

Think of Pac-Man, the old arcade game where you eat dots in a maze while four ghosts chase you. Each ghost follows a short rule in the code. The red one heads for the square you’re on; the pink one aims for a spot a few squares ahead of you. Hardly any player knows those rules. They know something else: “Red is always on my tail. Pink tries to cut me off.” That’s a mental model, and it’s good enough to play well with.

While someone plays, their hands press buttons and the machine shuffles numbers. In their head, something quite different is being built: a little world with things in it, actions they can take, and goals worth chasing. That world is the real product. The code is how you deliver it.

Stage magicians are experts at this. With timing, gestures and chatter, a magician gets a whole room to build a picture of what’s happening, and it’s carefully the wrong picture. The coin is in the left hand. Then the hand opens, the picture collapses, and that’s exactly what people paid for. A game designer uses the same skill more kindly, shaping what someone believes by choosing the evidence they see.

One step each
Bring the pie back to the door without Bob seeing you. Every time you take a step, he takes one too.

Patterns you bring with you

Mental models are built out of smaller parts called schemas. A schema is a pattern you’ve pulled out of lots of examples. A five-year-old can tell a dog from a cat, even a breed they’ve never seen, because they have a schema for each. You have one for restaurants too: wait to be seated, read the menu, order, eat, pay. It doesn’t always hold (try a food truck), but it tells you what to expect when you walk in.

Players arrive at your game carrying schemas from everywhere. In action games, a red barrel probably explodes. A cracked wall can probably be broken. Something that glows can probably be picked up. Plenty of adventure games put a splash of yellow paint on the ledges you’re allowed to climb, and once you’ve noticed it, you see it everywhere. Nobody learns these from a manual. They pile up, one game after another.

Games also build new schemas fast. Early in Super Mario Bros., where a plumber runs and jumps across a side-on world, a small brown mushroom creature walks straight at you. Touch it and you lose a life; jump on it and it’s squashed. One meeting is enough to know what to do with every one after it, and not a word of text was needed. Everything you show a player teaches them something like that, whether you meant it to or not.

Who gets in?
Creatures you’ve never met wander up one at a time. Say hi, or jump over them. Go with your gut.

Players are little scientists

Watch a child with a new toy. They poke it, shake it, drop it, and poke it again a bit differently. Now watch a family learning a new board game without reading the rules properly. It’s the same thing: someone tries a move, someone else says “I don’t think you can do that,” and they try something else.

That’s the scientific method, more or less, and we’ll call it the hypothesis loop: guess, try, look, adjust, repeat. Each trip round the loop adds to the model or fixes part of it. Your half-minute with the hotel remote was six or seven trips.

The Legend of Zelda: Breath of the Wild, an adventure game set in a huge open kingdom, is full of these trips. In a thunderstorm, lightning keeps striking close to you. You wonder whether it’s your metal sword, so you put it away, and the strikes stop. Now you know a rule nobody told you, and the next guess comes on its own: what if you dropped a metal sword next to an enemy? Players tried it, and it works.

There’s a neat flip side. Science takes a mountain of observations about the world and squeezes them into a few short rules, like the laws of physics. A game runs the other way. It starts with a few short rules and blows them up into a huge spread of things that can happen. The player, standing in the middle of it all, tries to squeeze it back down into rules they can carry around.

So everything that happens in your game is evidence. Clear, consistent evidence builds a good model quickly. Muddled evidence builds a confused player who decides the game is random. Even losing is evidence, and in Level 5 you’ll see how to design an interesting failure that sharpens the model without crushing the player.

The rune stones
Step on the plates to make the stones glow like the pattern in the corner. Nobody will tell you what the plates do.

True isn’t always right

You’d think the picture you build in a player’s head should be accurate. Often it shouldn’t, for a practical reason: player expectation, what the player already believes will happen, picked up from life, films and other games. When a game goes against a strong expectation, players rarely think “how interesting, I’ve learned something.” They think “this is broken.”

Films know this well. Cars that crash burst into flames. Explosions in space go boom, though there’s no air up there to carry the sound. None of it is true, and audiences would miss it if it were gone.

Platform games are full of polite fibs like this. In real life, once your feet leave the ground you can’t change direction; you go where you jumped. In most platformers you can steer in midair. Many also let you jump for a split second after you’ve run off a ledge, a trick designers call “coyote time,” after the cartoon coyote who hangs in the air until he notices the cliff has gone. Players almost never notice it. Take it out, and they say the controls feel broken.

The honest jump
Cross the jungle with the usual game jump. Then switch to “Real jump” and try it again.

Whether to be accurate depends on what the game is for. A flight simulator that trains real pilots has to be accurate, because the trainee must leave with a correct model of how the plane behaves. A game made for fun is free to take liberties. An educational game sits in between and has to pick its battles. You don’t have to side with expectation every time, or with truth every time. Decide on purpose, and know each place where your game agrees with what players expect and each place where it doesn’t.

From now on, describe your game twice. The first description is what it does: the rules, the pieces, the scoring. The second is what a brand-new player will think is going on. The two never quite match. Players will invent rules you never wrote, miss rules you thought were obvious, and bring in habits from every game they’ve played. You can’t see inside their heads, but you can predict, watch and compare, which is what playtesting is for (Level 18).

Try it · Pocket Game, step 3Five lines and a prediction

Write the rules of your Pocket Game in five lines or fewer, as if for a stranger. Then, on the back of the paper, write one thing (or three, if you’re on a roll) that you predict a new player will wrongly assume. Think about what your objects usually mean. Coins suggest money or heads-or-tails. A die suggests rolling and moving. If your game uses them differently, a new player’s schemas will fight you. For each prediction, decide: change the rule to match the expectation, or keep the rule and find a way to show the difference? Don’t show anyone yet. You’ll test these predictions in Level 18.

Carry forward: Design the picture in the player’s head, and choose on purpose where it agrees with what they already believe.

Check

Change Circle’s mind

Circle thinks red means danger. In this new game, red tokens are good. Pick the first three things Circle sees, so that Circle will happily grab a red token when the game starts.