The method, in full
Why we play everything twice
This page exists so that nothing about our process has to be taken on trust. If a review of ours ever contradicts what is written here, the review is wrong and we will correct it.
The first run is for the game. The second is for the review.
A first playthrough is dominated by novelty, and novelty is the least durable quality a game has. On a first pass everything unfamiliar reads as interesting, every locked door reads as promise, and every awkward system reads as something you have not understood yet. Some of those readings turn out to be correct. Most do not.
So the first run, for us, is simply playing. We take no structured notes and reach no conclusions. We finish it, put it down for at least a fortnight, and then start again on the hardest difficulty the game ships with — no mods, no assists we would not recommend, and no fast travel unless the game is unreasonable without it.
The second run is where the review comes from. The surprises are spent, so what remains is structure: pacing, repetition, whether encounters are designed or generated, whether the story survives knowing the ending. This is the version of the game a reader will be living in at hour thirty, and it is the version almost nobody reviews.
The hour log
During the second run we keep a paper sheet ruled into hours. One line per hour, written at the time, answering three things only: what changed, what was introduced, and what was repeated. The constraint is deliberate — a long note written mid-session is a note written badly, and a note written after the session is a memory rather than a record.
When the run ends, the sheet is transcribed and the two marks that appear on every review fall out of it almost mechanically.
The hour it clicked
The first hour where the log stops describing confusion and starts describing intent. Not the first hour we enjoyed — the first hour where the game's own logic became legible enough to play deliberately.
In our archive this ranges from hour two to hour fourteen.
The hour it stopped
The first of three consecutive hours where the log says nothing except more of the same. Three, not one, because every good game has a quiet hour and it would be dishonest to punish it for that.
One game in our archive has never reached this mark in eighty-three hours.
Minimum hours before publication
What we consider a finished second run
Genres end differently, so "finished" needs defining per genre rather than in the abstract. These are the thresholds we hold ourselves to. If a run falls short, the piece does not publish — that has happened four times, and two of those games we simply never covered.
Hours are cumulative across both runs and always stated in the review itself, so you can check our arithmetic.
| Genre | Second run counts as finished when | Minimum total |
|---|---|---|
| Action role-playing | Main path completed plus every optional region entered | 45 hours |
| Grand strategy | Two full campaigns, one played deliberately badly | 60 hours |
| Survival horror | Completed on the hardest setting with permadeath if offered | 15 hours |
| Racing simulation | Career completed plus ten laps of one circuit in the wet | 35 hours |
| Metroidvania | Completed without fast travel; map drawn from memory first | 25 hours |
| City building | Two settlements past their first crisis on hardest terrain | 50 hours |
Test hardware
Three machines, all of them ordinary
We do not own a flagship rig and we think that is an advantage. Every performance claim in a review is tied to one of these three machines by name, because "runs well" means nothing without knowing what it ran on.
| Name | Specification | Display | What it represents |
|---|---|---|---|
| The desk | Six-core desktop, 32 GB memory, mid-range 2023 graphics card | 1440p, 144 Hz | A comfortable but unexceptional enthusiast build |
| The kitchen laptop | Eight-core laptop, 16 GB memory, integrated graphics | 1080p, 60 Hz | What most readers actually own |
| The handheld | Portable x86 handheld, 16 GB shared memory | 800p, 60 Hz | Battery life and text legibility on a small screen |
Where a game is unplayable on the kitchen laptop or the handheld, we say so explicitly rather than omitting it. Two of the seven entries in our archive carry that warning.
Accessibility
Six things we check on every second run
These are logged whether or not they affect our own play, because a review that only describes the experience of the reviewer is a diary.
- Can every control be remapped, including on a controller?
- Can interface and subtitle text be scaled independently of each other?
- Is any critical information conveyed by colour alone?
- Can screen shake, motion blur and camera sway be switched off?
- Are timed inputs adjustable, or is there an alternative to them?
- Does the game remain playable with sound entirely off?
A page of the log from the racing entry. Hours forty-one to fifty-two, where the same wet circuit was run eleven times.
Awkward questions
Things readers challenge us on
Badly, if we are honest — there is no system. The two of us each nominate what we finished that quarter, and we cover whatever both of us are willing to sit through again. That biases the archive towards games we already suspected were good, which is a real limitation and worth knowing when you read us. We do not review games we abandoned on the first run, because we do not think a review from an unfinished run is worth publishing.
It is useless if you want to know whether to pre-order. It is the only thing that is useful if you want to know whether to buy the game that exists now, after three patches and a balance pass. There are plenty of places to read a launch-day verdict; we are deliberately not one of them.
No. Every game in the archive was bought at retail with our own money, on the same storefronts our readers use. We have declined early access four times, most recently in March 2026. It is not a moral position so much as a practical one: a pre-release build is not the game anyone will play.
We publish both readings. It has happened once, on the action role-playing entry from November 2025, and rather than averaging two opinions into a mush we left two closing sentences side by side. Readers told us it was the most useful thing we had run all year, which was mildly deflating and probably correct.
Repeatedly, and the corrections are listed with dates on the write in page. The most embarrassing was a performance claim tied to the wrong machine, which a reader spotted within a day. We do not silently edit published pieces; the fix and the reason both stay visible.