Third Act Review
Structure · pacing · plot
A quarterly of long-form criticism about how story-driven games are built, and about the point in the second half where a plot either holds or stops holding.
Third Act Review
Structure · pacing · plot
A quarterly of long-form criticism about how story-driven games are built, and about the point in the second half where a plot either holds or stops holding.
Articles here make measured claims about intervals, placement and density. Those measurements were made by one person with a notebook, and a reader is entitled to know exactly what was counted and how. What follows describes the procedure and its limits.
Games are bought at retail and played on ordinary consumer hardware. Default difficulty unless a game's structure changes on another setting, in which case the setting used is named in the article. Fast travel is used only where the game provides it and where a player would reasonably use it.
No article is written from a partial playthrough. Where a game has branching structure, the article addresses the path taken and states that a path was taken.
For the density measurements, an event is consequential when the plot's situation is different afterwards: an alliance changes, information becomes available that alters the objective, a place becomes accessible or closed, or a character's position shifts.
Combat encounters, item acquisitions and optional conversations are not counted unless they satisfy that test. This is a judgement call made case by case, and it is the largest source of imprecision in the figures.
| Measured | Definition used |
|---|---|
| Consequential event | Plot situation differs afterwards |
| Interval | Playtime between consecutive consequential events |
| Central turn | The revelation the plot is organised around |
| Threatening event | Anything the game frames as danger |
Against in-game playtime as displayed by the game or the platform, recorded at the moment the event occurs. Where a game does not display playtime, a stopwatch is used and the session start is noted. Loading screens are included in the interval because the player experiences them.
Figures are rounded to the nearest five minutes in the text, since the precision of the underlying measurement does not justify anything finer.
By whether their structure raises a question that several thousand words can address, not by quality or release date. A game with an instructive failure is more useful to this publication than a game that works and cannot be taken apart.
Sample sizes are small, deliberately stated, and never generalised. Six of eight means six of the eight I played and measured, and nothing about the medium.
Scores, recommendations, purchase advice, news, previews, patch coverage and release-day reviews. An article concluding that a game's middle sags is not advising against playing it, and nothing here should be read as guidance about how to spend money.
The definition of a consequential event is mine and another person applying it would produce slightly different numbers. The samples are small. My attention varies across a forty-hour playthrough, and a quiet stretch that felt long at two in the morning may have been ordinary.
I also cannot know why any structural decision was made. Where an article attributes a pattern to scheduling or budget, that is an inference from the finished work and from publicly available accounts of production, not a report of anything I have been told.
Miscounts are corrected in place and the correction is stated. Interpretations are not revised because a reader disagreed. Where a patch changes a game's structure after publication, the article stands as a record of what the game was and a new note describes what it became.
Writing about structure means describing where things happen, which includes endings. Every article here does so without internal warnings, and readers who prefer to arrive unprepared should read afterwards.