We played all nine to completion, recorded every line of the final sequences and compared them against each other. Six differ only in a closing paragraph. Three are genuinely distinct in what happens and what it means. A game is entitled to that structure; a store page is not entitled to describe it as nine.
Counting endings is harder than it sounds, which is why we published a definition before we started doing it. An ending counts as distinct if it differs from every other ending by more than a single closing scene, meaning something happens differently rather than merely being described differently.
By that standard Nine Doors has three endings. Six of the nine share an identical final sequence and diverge only in a concluding paragraph that names a different character. This is not a bug or an accident; it is a reasonable way to acknowledge a player's choices cheaply. It is only a problem because the number nine is doing promotional work the content cannot support.
We recorded each ending in full and compared them for divergence in the final sequence, in the events preceding it, and in the state of the world afterwards. The groupings below are ours and we have described the method on the standards page so that a reader can disagree with them precisely.
| Ending | Final sequence | Divergence point | Group |
|---|---|---|---|
| A | Unique, 11 minutes | Chapter 4 | Distinct 1 |
| B | Unique, 9 minutes | Chapter 5 | Distinct 2 |
| C | Unique, 14 minutes | Chapter 3 | Distinct 3 |
| D | Shared with E, F, G, H, I | Final scene only | Variant of 2 |
| E | Shared, one paragraph differs | Final scene only | Variant of 2 |
| F | Shared, one paragraph differs | Final scene only | Variant of 2 |
| G | Shared, one paragraph differs | Final scene only | Variant of 2 |
| H | Shared, one paragraph differs | Final scene only | Variant of 2 |
| I | Shared, one paragraph differs | Final scene only | Variant of 2 |
Reaching all nine took us nineteen hours, of which we estimate four were new material. That ratio is the actual complaint. Had the game presented three endings and six epilogue variations, we would have played it in seven hours and thought considerably better of it.
Chapter six contains a nineteen-minute stretch with no reachable checkpoint and no manual save. We lost it twice: once to a crash and once to closing the game while believing an autosave had occurred.
A game asking to be replayed nine times has an unusually high obligation to make replaying convenient, and this one does the opposite. There is a single save slot. Reaching a second ending means overwriting your route to the first, so any comparison requires either a full replay or manual file copying outside the game.
We did the latter, because our method requires it. A reader who wants to see two endings without nineteen hours of effort will have to do the same, and being obliged to manage save files in a file browser to experience a game's advertised content is a design failure regardless of how the endings are counted.
| Quit point | Work lost | Note |
|---|---|---|
| Mid-conversation, chapter 2 | 3 minutes | Acceptable |
| Mid-puzzle, chapter 4 | 7 minutes | Puzzle state not preserved |
| Chapter 6 traversal | 19 minutes | No checkpoint in the section |
| Immediately before an ending | 11 minutes | Final sequence restarts entirely |
A game that wants nine playthroughs owes you nine save slots. This one has one, and expects a file browser to make up the difference.
The three distinct endings are strong, and the earliest divergence at chapter three is brave: it commits to a substantially different second half rather than a different epilogue. Ending C in particular reframes two earlier scenes convincingly enough that we went back to check them, which is the reaction a branching structure should aim for.
The writing throughout is careful and the mid-game pacing is good. Our objections are structural rather than aesthetic, and a reader who never intends to see more than one ending will encounter a much better game than our score suggests.
Three well-made endings buried inside a claim of nine, with nineteen hours required to see roughly four hours of unique material. One save slot for a game built on branching, and a nineteen-minute checkpoint gap in chapter six that cost us the same stretch twice. The writing deserves a better structure than it has been given.
