Skip to content

How we score

The ten-point scale, what the four subscores mean, how a reader score differs from an editorial one, and the rules a review has to follow before it is published.

One scale, everything on it

A game, a novel, a record and a graphics card are scored on the same ten points, to one decimal. The number is anchored to the reception a work actually had, then moved by our own judgement of it. The catalogue is curated, so it skews high; the tiers below are what the number means in practice.

ScoreTierWhat it means
9.5–10MasterpieceA landmark of its medium. About one entry in twenty.
9.0–9.4EssentialExceptional; the flaws are footnotes.
8.0–8.9ExcellentExcellent, with flaws a reader should know about.
7.0–7.9GoodGood; the reservations are real.
6.0–6.9MixedSome will love it; say who.
Below 6PoorGenuinely flawed. Used when deserved.

Four subscores, not a formula

Every review carries four subscores suited to its medium. They are judgements in their own right rather than a calculation: the overall score is not their average, though it stays within a point of it and has to be consistent with them.

Games
Gameplay · Story · Presentation · Longevity
Movies
Story · Direction · Performances · Craft
TV series
Story · Characters · Performances · Craft
Anime
Story · Characters · Animation · Sound
Books
Writing · Story · Characters · Ideas
Comics
Story · Art · Characters · Pacing
Albums
Songwriting · Production · Performance · Cohesion
Songs
Songwriting · Production · Performance · Replay value
Hardware and devices
Performance · Design · Features · Value

Two scores that never mix

The editorial score is ours: one per work, from the review. The reader score is the mean of every rating readers have given, shown with the number of ratings behind it so you can judge how much it is worth. They are displayed separately and never averaged together.

What a review may and may not say

  • No claims of first-hand testing. We do not say we played for forty hours or measured a battery. Where performance matters, we report what was widely reported and what the specification implies, and say which it is.
  • No invented numbers. No frame rates, sales figures or award counts unless they are certain; when they are not, the point is made qualitatively or not at all.
  • Premise and first act only. No endings, twists or late reveals.
  • Real weaknesses, named. Every review lists what is wrong with the work, not only what is right.
  • Sources are listed. Facts come from official APIs and public datasets, never from scraping, and each entry shows where its data came from.

When a fact turns out to be wrong, the entry is corrected and the correction is recorded with the batch the entry arrived in.

Reader reviews

Readers can review anything in the catalogue. Every submission is read by a moderator before it appears, and the score a reader gives counts towards that work’s reader average.