What this publication is
One person, playing PC games in the last hour or two of the day, writing down when each session actually ended. That is the whole apparatus. There is no team, no office and no funding beyond the games being bought at retail.
The premise is that reviews are almost universally written from the perspective of someone with time available, and that a very large number of people play games in the narrow window between finishing work and needing to sleep. In that window a game's quality is not the only question. Whether it will let you leave is the other one, and nobody measures it.
The intended session, declared first
Before each evening's play, an intended session length is written down. This is the single most important part of the procedure and the part most easily got wrong: it must be decided before the game launches, not adjusted afterwards to match what happened.
When the session ends, the actual length is recorded beside it. The difference between the two figures, averaged across at least twenty evenings, is the overrun — and the overrun is a far better description of a game's demands than any total playtime figure.
Twenty evenings minimum
No figure is published on fewer than twenty separate evenings, and several games in the file have thirty or more. Session length varies enormously with how tired the player is, how the day went, and whether anything else is happening, and a small sample produces numbers that describe a mood rather than a game.
This requirement is why thirty-one games have taken as long to measure as they have, and why six of them have reviews while twenty-five appear only as figures in the logs.
| What is recorded | How |
|---|---|
| Intended session | Written down before launch, never revised |
| Actual session | Clock noted at the moment play stops |
| Time to first input | Cold start to controlling something, by stopwatch |
| Safe exit points | Counted per session, from actual attempts to leave |
| Loud events | Tallied whenever audio spikes above the ambient bed |
| Stop quality | Assigned from what leaving actually cost, not from the options menu |
The four stop grades
Clean. A session can end at a natural boundary and nothing is lost. Every clean game in the file has a smallest unit of progress under six minutes.
Soft. Leaving costs a small amount of repeated play, under five minutes. Common and largely forgivable.
Sticky. Progress commits only at a milestone, and the milestone's arrival cannot be predicted in advance. This is the grade that produces the largest overruns.
Hostile. The game imposes a penalty for leaving — a lost map, a broken streak, a timer that ignores the pause menu. Six games in the file are graded this way and one of them is reviewed.
The grade describes what leaving costs, measured from having actually tried to leave, rather than what the save system claims in a settings screen.
What the score is separately
The score out of ten judges the game as a game, played across whatever number of hours the byline states. It is not weighted for lateness in any way. The highest-scoring game here happens also to have the cleanest exit, and the second-highest scoring game in the wider file is graded hostile.
Keeping the two judgements apart is deliberate. Collapsing them into a single number would produce something that tells a reader neither whether a game is good nor whether it suits their evening.
What is refused
No review copies, no press access, no early builds, no embargo agreements, no sponsored coverage, no affiliate links. Every game is bought.
This is a practical requirement rather than a stance. Measuring twenty evenings takes a month, which is incompatible with review-window scheduling, and a game supplied on condition of timing cannot be measured this way at all.
The known weaknesses
One player, one household, one machine. Session length is affected by circumstances that have nothing to do with the game, and averaging across twenty evenings reduces that noise without eliminating it.
The sample is also weighted towards single-player games. Eight of thirty-one involve other people, and none of the four grades properly describes a situation in which leaving costs somebody else rather than you. A fifth grade is likely and has not yet been defined.
Finally, patches change games. Where an update alters a save system the old figures are struck out and the game is measured again from scratch over twenty fresh evenings, which has happened three times.
In short
Declare the intended session, record what actually happened, repeat for twenty evenings, grade the exit from real attempts to leave, publish both figures separately, and buy every game.