What it is
A conversation across one kitchen table over the course of a single evening, in which two people work out what to say to each other about something that happened eleven years ago. You control which of them speaks, and the choices are not branching in the usual sense — they set the temperature of the room rather than the plot.
Four hours. No inventory, no map, no menu deeper than two levels. It is impossible to be lost, impossible to be under-levelled, and impossible to arrive back after a break unsure what you were doing, because the thing you were doing is a conversation and the last two lines of it are on screen.
Hours to a real ending
Four hours and eleven minutes on a first playthrough, with no guides and no skipping. Second playthroughs run around three hours because the reading is faster, and the game does not pretend otherwise: there is no artificial gate on a second run and nothing hidden behind repetition.
What is unusual is the internal shape. There is no lull. The audit logs a dip in engagement in almost every game it examines, usually at the point the design runs out of new ideas and starts repeating. Here the material is exhausted at exactly the moment the game ends, which is either careful editing or the smallest scope in the ledger. Probably both.
| Segment | Minutes | Note |
|---|---|---|
| Opening | 0–38 | Establishes the room, the pair, and what is being avoided |
| First turn | 38–104 | The subject arrives; the best writing in the game |
| Long middle | 104–201 | Slower, deliberately uncomfortable, holds its nerve |
| Close | 201–251 | Resolves without tidying; no epilogue, correctly |
Cost of returning after a break
Two minutes. Put down for three weeks at the ninety-minute mark and resumed cold, the entire cost of re-entry was reading the two lines of dialogue still on screen and remembering which of the two people I had been steering towards honesty.
This is not an achievement of clever design so much as an achievement of restraint. There is no state to reconstruct because the game holds almost no state. Fourteen ledgers of data suggest that this single property — how much of the game exists only in the player's memory — predicts abandonment better than length, genre or difficulty.
The lesson is not that games should be four hours long. It is that a game which stores its own context on screen can be put down safely, and a game that stores it in your head cannot.
Where owners stop
Almost nowhere, which makes this the least interesting section of the audit. The only measurable hesitation is at around thirty-eight minutes, where the game's subject arrives abruptly and some players will decide it is not what they want to spend an evening on. That is a taste boundary rather than a design cliff, and it is arguably a strength.
What it does that longer games could copy cheaply
- Keeps the current thread visible on screen at all times, in plain sentences.
- Uses no proper nouns the player is expected to memorise.
- Has no build, loadout or configuration to reconstruct.
- Ends when the material ends rather than when the runtime target is hit.
Verdict
The highest finish estimate in fourteen ledgers, and not because it is small enough to be trivial. It is the clearest demonstration the bureau has of a game designed for a player with a life. If you buy one thing out of this ledger and want to actually see the end of it, buy this.
It is also the game that makes Cloudstair Ascent look worse than it deserves, which is a distortion worth naming. Comparing a four-hour conversation to a two-hundred-hour climb on finish rate alone is not a fair fight, and the method page says so at length.