Lesson 13 · Part 3: The Machinery
Loops, Snowballs, and Chaos
Why some games decide themselves too early, how loops make a world speed up or settle down, and a few simple pictures for thinking about anything that moves.
Loops, Snowballs, and Chaos
It’s recess, and somebody has brought marbles. You’re playing “keepsies”: whatever you knock out of the ring, you get to keep. One kid is a little better than everyone else. She wins a few marbles early. Now she’s got more marbles to shoot with, so she wins more, and now she’s got even more. By the end of the week she owns a shoebox full of everybody’s marbles, and nobody will play with her anymore.
Nobody cheated. The rules were simple and fair. But they had a shape built into them, where winning made it easier to win again, and that shape quietly ruined the game. This level is about spotting shapes like that before they sneak into yours.
The thermostat and the snowball
Start in the shower. It’s cold, so you turn the handle toward hot. Nothing happens for a couple of seconds, so you turn it further, and now you’re being boiled. You yank it back, overshoot, freeze, and after three or four goes you find the spot. For that half minute you were part of a little machine: you felt the temperature, compared it to what you wanted, adjusted, and felt again.
That’s a feedback loop, where the result of an action loops back around to shape the next action. A thermostat runs one all day long. Too cold, heat on. Warm enough, heat off. The old steam engines had a lovely version called a governor: two spinning weights that flew outward as the engine sped up and, by flying outward, closed the steam valve a little. The faster the engine ran, the harder the governor held it back.
Thermostats and governors are balancing loops. They pull things back toward a target and fight anything that pushes them away. There’s an opposite kind, too. A snowball rolling downhill picks up snow, which makes it bigger, which makes it pick up snow faster. A microphone too close to its own speaker goes from a hum to a shriek in about a second. Those are reinforcing loops: they push a system further in whatever direction it’s already going, faster and faster.
Keepsies is a reinforcing loop wearing a playground disguise. Winning gives you marbles, marbles help you win.
The runaway leader
Every game designer eventually meets the runaway leader. You’ve probably met them too. Think of that long family board game about buying up streets and charging rent. The player who gets ahead owns more property, so they collect more rent, so they can buy even more property. About halfway through, everyone at the table knows who’s going to win. Then you play for another hour and a half anyway.
That’s a reinforcing loop with nothing to stop it. And here’s the twist: snowballs aren’t bad. A reinforcing loop is what makes early decisions matter. If nothing you do in the first ten minutes counts for anything later, why bother? Snowballs also give a game an ending. Things build up, tip over, and resolve.
The trouble is when the snowball rolls too early or too hard, and the game gets decided long before it’s over. So designers reach for balancing loops, usually called catch-up rules. You know a lot of them already:
- Handicaps. In golf, weaker players get extra strokes so two friends of different skill can have a real match.
- Staggered starts. Little kids get a head start in the family race at the picnic.
- The loser goes first. Or chooses the next board, or deals.
- Gang up on the leader. In plenty of card and party games, the other players naturally team up against whoever’s ahead. Nobody wrote that rule. It grows out of the game.
- Bonuses for the back of the pack. Some racing video games quietly hand better bonus items to the drivers at the back, to keep everyone in it.
Catch-up rules have a cost, too. Lean on them too hard and being good stops mattering, and winning starts to feel like a coin toss, which damages fairness in a different way. Too much snowball and the ending is boring. Too much catch-up and the beginning is pointless. Getting the mix right is a big part of what designers mean by balance. Most beginners build one kind of loop or the other by accident. Learn to spot which one you’ve got.
The bathtub
Back in the 1950s, an engineer named Jay Forrester came up with a way to model really big things, like a company or a whole city, and called it system dynamics. It’s basically plumbing. There are three kinds of part: containers that hold an amount of something, flows that carry stuff in and out, and valves that set how fast each flow runs.
Picture a bathtub. The tap is a flow in, the drain is a flow out, and the water level is what’s in the container. Now relabel it. The tub is a town’s population. The tap is people being born, the drain is people dying, and you can add a second tap for people moving in.
The fun starts when a valve depends on a level. Say fewer people move to a town as it gets crowded. Then the fuller the tub, the more the tap closes. Congratulations, you’ve built a balancing loop out of plumbing. Say instead that a busy town attracts even more people. Snowball.
You can use this right away. Every number in a game that goes up and down is a bathtub: points, money, cards in hand, the tokens in the middle of the table. For each one, ask what fills it, what drains it, and what controls the taps. Forrester’s work was a big influence on SimCity, a computer game about running a city, which joined his tubs-and-pipes thinking to a cellular automaton like the ones in Level 12, a grid of neighborhoods, so the numbers had somewhere to live.
That’s a good moment to notice something. A grid and a bathtub are two different ways of looking at the same city. Each way of modeling a system, each lens, is called a paradigm. No lens is the “true” one. Each brings some things into focus and blurs others, and you can mix them.
The marble and the bowl
Here’s a picture for something even stranger. Take a salad bowl and drop a marble in from anywhere near the middle. It rolls around and settles at the bottom, every time. That’s a stable system: where it starts barely matters to where it ends up.
Now turn the bowl upside down and drop the marble on top. Land a hair to the left and it rolls off to the left. A hair to the right and it ends up under the sofa. That’s chaos: a tiny difference at the start grows into a huge difference at the end. A chaotic system follows its rules exactly. Like Life in Level 12, it’s a matter of determinism, no dice involved. It’s unpredictable in practice because you’d need to know the starting point perfectly, and nobody ever does. Weather works this way. That’s why forecasts get shaky after a few days.
People found this out the hard way. Early computer models of the whole world’s economy and food supply made confident forecasts, and when a couple of their inputs turned out to be slightly off, the results came out wildly different. That shock was one of the first hints of what became chaos theory.
Two lessons for a designer. First, you can’t fully predict your own game on paper, so you have to play it. Second, every game sits somewhere between order and disorder, and you get to choose where. Very orderly games are dependable and eventually dull. Very chaotic ones are surprising and eventually feel random. A lot of good games keep the core predictable and let a little chaos into the edges.
The rumor
One more picture: a rumor going around a school. It travels along friendships, and a network of friendships behaves in ways that are hard to guess.
Start with a crowd of strangers and introduce people at random. For ages nothing much happens. You get little clumps of two or three, and a rumor started in one clump dies there. Then, at around one link per person on average, most of the crowd suddenly snaps together into one big connected cluster. No single introduction causes it, which makes it a small case of emergence. Once that cluster exists, the paths across it are surprisingly short. The popular figure is “six degrees of separation”: a friend of a friend of a friend, six times over, is supposed to reach nearly anyone on Earth. Whatever the exact number, the idea holds. Big networks are much smaller across than they look. That’s the small world effect.
For a designer, anything that spreads through your game (a disease, a fashion, news of what the player did) travels on a network, and it’ll tip suddenly, not gradually.
Two dials worth carrying
Two more ways to look at any system, like dials on a panel.
The first is local and global: how far does an effect reach? One noisy classroom is local. The school bell is global. In a city game, a traffic jam is one street, and the tax rate is the whole city. Players need both. Local things give them something to fix right now and see the result. Global things give them a reason to think about the whole.
The second is cooperation and competition. Level 1 introduced the zero-sum game as a choice. It’s more useful to think of it as a dial, with lots of positions between “all together” and “all against,” and the dial can move during play. Tug-of-war is cooperation inside each team and competition between them. An old computer game called M.U.L.E., about settlers on a new planet, nailed this: everyone’s trying to get richest, but if nobody helps anybody, the whole colony collapses and everybody loses.
Try it · Pocket Game, step 13The runaway leader
Play a round of your Pocket Game and watch whoever gets ahead. Find one loop that helps the leader, where being ahead makes it easier to get further ahead. Then find one that helps the player behind. If you can’t find one of them, that’s your answer: your game only has one kind. Now adjust one loop. Weaken the leader’s advantage (the winner of a round goes second next time) or add a small catch-up rule (the player behind gets one extra try). Play again and notice whether the game stayed close for longer.
Carry forward: For anything in your game that grows or shrinks, ask what fills it, what drains it, and whether winning makes winning easier.