Everything below applies to every report in the library without exception. Where a report deviates, the report is wrong and we will correct it.
A game is played to completion blind, at default difficulty, with no guides, patch notes or community resources consulted. It is then played to completion again with full knowledge, timed per section. Both hour counts appear on the report, because the difference between them is frequently more informative than the score.
Neither run may be abandoned. If a reviewer cannot complete the second pass, the game does not receive a score and we publish the notes with an explanation instead.
Between the second run and the write-up, three checks are performed. The game is played on its hardest setting far enough to establish whether difficulty is a change in behaviour or merely in numbers. It is run on the weakest machine we keep for the purpose and the results are described in specifics rather than adjectives. And the save file is examined for size and load times after extended play, because a game that degrades at hour sixty is a different product from the one you buy.
Reports are written seven days after the second run finishes, from notes, without the game running. This is deliberate: it filters out novelty, which is the single largest distortion in games criticism. What a reviewer can still articulate a week later is what actually stayed.
One number from one to ten with a single decimal. No sub-scores, no weighted categories, no averaging. The number answers one question: how well does this game hold up once you know everything it was hiding?
Nothing. Value depends on how much time a reader has and what they intend to do with it, neither of which we know. Where cost changes our practical recommendation, the report says so in a sentence rather than adjusting the number.
Games without an ending cannot be played twice in a way this method can assess, so live-service titles, endless simulations and competitive multiplayer receive written coverage without a number. Early access builds and pre-release versions are described with the exact build named and are never scored. Games we abandoned are published as notes with the reason stated.
A score is reopened when a patch changes something the report criticised, when a later update removes something it praised, or when we conclude we were mistaken. In every case the original number stays on the page next to the new one, with the date and a paragraph of reasoning. Six scores have been revised across forty-two reports: four upward, one downward, one because a reader demonstrated we had misread a system.
Where a copy was supplied by its maker, the report says so in the opening block. Most of what we play is bought at retail. No report is sponsored, no coverage is arranged in exchange for access, and this site carries no advertising, affiliate links or paid placements. Nothing is for sale here.
Reviewers do not cover games made by people they have worked with or know personally. Where that overlap exists, another contributor takes the report and the relationship is disclosed on it.
A method built on second playthroughs cannot avoid discussing endings, and pretending otherwise would make the reports useless. Where a report intends to describe a late-game event, it warns you in the paragraph before, and we describe the shape of a conclusion rather than its content wherever the argument allows.
Factual errors are corrected in place with a visible note stating what changed. Hour counts are never quietly adjusted; if a figure was wrong, the correction says what it was and what it should have been. We consider silent edits worse than the original error.
Quote us in discussion and commentary with attribution. Scores may not be used in advertising or on store pages, because a number extracted from a two-pass report misrepresents the only thing that makes it meaningful.
This site collects nothing. No accounts, no comments, no newsletter, no forms, no analytics and no advertising technology on any page. The privacy notice sets that out in detail.
