The classification rule is simple and applied at the moment of writing rather than afterwards. If an event happened because the game did something — a season turned, a bank failed, a settler left — it is supplied. If it happened because I decided something the game had not proposed, it is invented. Borderline cases are recorded as supplied, which biases the figures downward and is stated here so the bias is visible.
The position this ledger started from
That a high invented figure indicates a generous sandbox. The reasoning seemed obvious: a game that leaves room for the player to author their own experience is doing the thing sandboxes exist to do, so more invention is better design.
Six notebooks in, the figures did not cooperate.
| Notebook | Invented | Hours logged | Closed because |
|---|---|---|---|
| Chalk Flats | 84% | 4 | Materials ran out |
| Quarry Gap | 81% | 52 | Exhausted the physics |
| Cog And Beam | 77% | 23 | Exhausted the materials |
| Plateau Holding | 62% | 96 | Satisfied |
| Terrace Works | 58% | 34 | Save lost |
| Moor Tower | 56% | 71 | Patch broke it |
| Spiral Station | 54% | 38 | Bored |
| Drift Habitat | 52% | 16 | Patch reset layout |
| Lowfield | 49% | 44 | Consequence |
| Orchard Row | 44% | 19 | Too forgiving |
| Salt Marsh | 39% | 27 | Mid-game wall |
| Pine Hollow | 31% | 11 | Nothing at stake |
| Ridge Claim | 29% | 8 | Save corrupted |
| Birch Lake | 26% | 29 | Nothing left to want |
The three highest invented figures in the file belong to games that ran to four, fifty-two and twenty-three hours. The three lowest belong to games that ran to twenty-nine, eight and eleven. Both ends are short. Everything long sits between forty-nine and sixty-two per cent.
The position it holds now
That the invented figure measures how much a game withholds, and that both extremes are failures of different kinds.
A game at eighty per cent invented has handed you a toolkit and stepped back. Nothing it does will surprise you, because it does almost nothing; the only novelty available is your own increasing competence, and competence saturates. Quarry Gap is the clean case — an outstanding physics solver that finished the moment I could predict it.
A game at thirty per cent invented has supplied so much that there is nothing left to add. Birch Lake is the clean case there: abundant materials, no failure state, and an invented figure made up almost entirely of restrictions I imposed on myself to manufacture difficulty the game would not provide.
The games we played longest are the ones that kept changing the conditions under which our own inventions had to survive. Fifty per cent is not a target; it is what a working conversation between a game and a player happens to look like when you count it.
What changed it
The Quarry Gap entry, specifically the section on why forty-one machines produced no accumulation. Up to that point every high-invented game in the file had also been short, and I had been explaining that away — Chalk Flats was too small, Cog And Beam was too limited. Quarry Gap was neither. It was an excellent, deep, generous physics sandbox with the second-highest invented figure in the file, and it still ran out.
What it lacked was any mechanism by which my own creations changed the situation for my later creations. Nothing persisted, nothing degraded, nothing constrained. Compare Plateau Holding, where a badly-dug channel at hour one was still dictating decisions at hour ninety.
The refinement that follows
The useful variable is not how much the player invents. It is whether the game converts invention into new conditions.
| Game converts invention into… | Notebooks | Median hours |
|---|---|---|
| New conditions that persist | 4 | 62 |
| Reversible arrangements | 6 | 21 |
| Nothing at all | 4 | 14 |
Four notebooks in the top band, which is far too few to be confident about. But the four are the four longest logs in the file, and the mechanism is legible in each: silt in Plateau Holding, structural load in Moor Tower, soil and water in Lowfield, and terrace drainage in Terrace Works.
Known weaknesses in the count
Memorability is judged by me, at the time, with no external check. An event enters the ledger because it seemed worth writing down, and I am certain I under-record quiet supplied events — a season passing without incident is not memorable but it is doing work.
The borderline rule also matters more than it should. Classifying ambiguous cases as supplied pushes every figure down, so the true invented percentages are probably a few points higher across the board. That does not affect the ordering, which is what the findings rest on.
And the sample is one person. A player who arrives at a physics sandbox with a project already in mind would generate a higher invented figure in Quarry Gap and might well play it for two hundred hours, which would leave this ledger describing my habits rather than the game.
What would reverse it again
A game above seventy-five per cent invented that holds for eighty hours or more. That single case would break the current position, and I have been looking for it since the Quarry Gap entry was written.
A game below thirty-five per cent invented with a long log would be almost as damaging — it would suggest that supplied content can carry a sandbox indefinitely, which the current position denies. Birch Lake is the closest the file has and it stopped at twenty-nine hours.
Current state of the ledger
412 events across fourteen notebooks. Invented figures range from twenty-six to eighty-four per cent, and every notebook over forty hours sits between forty-nine and sixty-two. The ledger no longer treats a high invented figure as a virtue: it now argues the useful property is whether a game converts player invention into persistent conditions, which four of fourteen games do and which those four notebooks are all long. Open, one person, and reversible on a single counterexample.