A commission for a four-year-old and a commission for their father are frequently the same game. Same characters, same artwork, same story. What differs is five numbers.
That is worth stating plainly because the instinct is to build two versions, and two versions is both twice the work and a worse result. The child's version ends up thin, and the adult's ends up excluding the child entirely.
The five numbers
Our platformer engine has three difficulty presets. Every one is a set of the same five values.
| | gravity | jump | speed | late-jump window | lives | | --- | --- | --- | --- | --- | --- | | Young child | 0.42 | 12.2 | 3.4 | 14 frames | unlimited | | Standard | 0.58 | 11.6 | 4.0 | 8 frames | unlimited | | Spicy | 0.72 | 11.2 | 4.6 | 4 frames | 3 |
Read across the top row and you can see the whole design philosophy for a small child.
Gravity is low, so a fall is slow and legible. A four-year-old watching a character drop needs time to understand what is happening. Fast gravity feels crisp to an adult and simply looks like disappearing to a child.
The jump is high, so reaching something is forgiving. Not because precision is impossible at four, but because failing to reach something you obviously meant to reach is the single most frustrating experience in a platformer.
The walk is slow. This is the one people get wrong. The instinct is to make a child's character fast and bouncy. In practice a fast character outruns a small child's reactions and they spend the game overshooting.
Lives are unlimited, and a fall returns you to the start of the level. The worst thing that can happen is a short walk back.
Coyote time, which nobody notices and everybody needs
The fourth number deserves its own explanation because it is invisible when it works and the game feels broken when it is missing.
When a character walks off the edge of a platform, there is a brief window afterwards during which pressing jump still works, as though they were still on solid ground. It is named after the cartoon coyote who runs past the cliff edge and only falls once he notices.
Human beings do not press jump at the exact frame their feet leave a ledge. They press it a fraction late, having seen the edge coming. Without a forgiveness window, every one of those presses fails, and the player concludes the controls are unresponsive rather than that their timing was two hundredths of a second out.
Our easiest setting allows fourteen frames, near a quarter of a second, which is very generous. The hardest allows four. That single number is a large part of why one game feels kind and the other feels sharp.
What must not be adjustable
The important half of the design is what stays fixed.
The level does not change. Same platforms, same collectibles, same route. If the easy version removes obstacles, then the child and the adult are not playing the same game and cannot talk about it. The entire point of a family commission is that they can.
The story does not change. No abbreviated version, no skipped cutscenes.
Nothing announces the difficulty. There is no "baby mode" label anywhere on screen. A seven-year-old who spots that they have been put on the easy setting will insist on the hard one and then have a miserable time.
And one absolute rule for a young child's build: there is deliberately no way to lose. Not "it is hard to lose". None. Falling off the bottom returns you to the start of the level and that is the entire failure state. A five-year-old who is given a game where they can be told they have failed will often simply stop playing it, and a gift that gets played once is a failed gift.
The same problem in a different genre
This is a general principle rather than a platformer one, and our chess build makes the same decision from the opposite direction.
There, difficulty is how far ahead the opponent looks, how long it may think, and how often it deliberately picks a weaker move. But the underlying design question is identical: what should losing to this feel like? The answer in both cases is that a beginner should feel like they are getting somewhere, not that they are being handled.
We wrote that one up separately in how the computer decides its move. The engineering is completely different; the reasoning is the same.
Proving it is actually possible
There is an obvious failure mode in all of this: setting the numbers kindly and then designing a level the settings cannot rescue.
It happened to us. Our example level required precise vertical platforming to reach the goal, and no amount of generous gravity made that appropriate for a four-year-old. The build was green, the game ran, and it was not completable by its intended player.
So completion is now tested automatically on every build. A machine plays the level at the target difficulty and the build fails if the goal cannot be reached. That check is described in proving a game is finished, and it exists precisely because a difficulty setting is a promise the level has to keep.
What to ask for in a brief
Say who will actually hold the controller, and say it in ages. "For the family" is not enough; a four-year-old and a nine-year-old are different games.
Say whether anyone should be able to lose. This is the single biggest tuning decision and it is a parent's call, not ours.
Expect to be asked about the youngest player. They set the floor. Everyone older is accommodated by a setting; the youngest is accommodated by the design.
The build diary covers a build made for a grandfather rather than a child, and turning a teddy into a game is the closer fit if the player is small. Two finished builds are in the arcade.
