Frame Rate FilesThe testing standard
Every game covered here is played on two machines: a current mid-range desktop and a six-year-old build that represents what a large share of players actually own. Both run Windows on standard drivers, with no vendor-specific tuning applied. Frame rate figures quoted in a review are sampled over sustained play rather than taken from a benchmark scene, because a benchmark scene tells you almost nothing about how a game behaves in its worst area at hour twenty.
Play logs are mandatory. Before a review is written, the writer records hours played, difficulty setting, patch version, number of crashes, and the moment at which the game either revealed its best idea or ran out of them. A condensed version of that log appears on every review page, because a score without a play count is a claim without evidence.
The scoring scale
Scores run from one to ten with one decimal place. There is no weighted formula, no category averaging and no numeric floor. The number summarises a judgement; the review contains the reasoning.
| Band | What it means | What to do |
|---|---|---|
| 9.0 – 10 | Does something the genre has not managed before, and does it well enough that the flaws are footnotes. | Buy it. |
| 8.0 – 8.9 | Confident, well-built, with weaknesses we can describe in one sentence each. | Buy it if the genre interests you. |
| 7.0 – 7.9 | One outstanding dimension surrounded by uneven work. Often a great twenty hours in a forty-hour shell. | Buy for the specific strength named in the review. |
| 6.0 – 6.9 | Real ideas, compromised execution, usually through scope or pacing. | Wait for a discount. |
| Below 6.0 | Broken, hollow, or hostile to the player. The review specifies which. | Skip it. |
Price is deliberately excluded from the score. Value depends on how much you play, which is information we do not have. Where price genuinely changes the recommendation, the review says so in words.
When a score changes
Five published scores have moved. A number is reopened when a content patch materially alters something the original review criticised, when a technical update fixes performance we called out, or when we simply got it wrong. In all three cases the original score stays visible next to the new one with the date of the revision and a paragraph explaining what shifted. Silent edits are not acceptable.
A publication that never revises a score is either extraordinarily lucky or is not paying attention to what happens to games after launch week.
What we do not score
Three categories get written coverage without a number. Early access and open beta builds are described with an explicit build version, because scoring something that is openly unfinished misleads readers and pressures studios. Games abandoned mid-run are published as a play log explaining why we stopped. Live-service titles are covered per season, with each article naming the season rather than pretending to judge something that changes monthly.
Guide standards
Every figure in a guide is measured in a live save on a named patch. Nothing is copied from a wiki, a video or a forum thread without testing it first, and where a widely repeated community number turns out to be wrong, we publish the correction with the measurement conditions attached. Guides carry a re-verification date, and when a patch invalidates one we either update it or mark it obsolete rather than leaving it quietly wrong.
Independence
No sponsored articles, no paid placements, no affiliate revenue tied to scores, and no coverage arranged in exchange for access. Review keys are accepted and disclosed on the review that uses them; roughly two thirds of what we cover is bought at retail because no key was offered. If a writer has any relationship with a studio, they do not review its games, and that relationship is disclosed on whatever coverage somebody else writes.
Corrections
Factual errors are corrected with a visible note at the point of the error, not silently overwritten. Three guides and two reviews currently carry correction notes. We consider that a reasonable number for a desk that publishes measured figures, and we would be suspicious of a site that claimed zero.