| Critics | Players | Signals | ||||||
|---|---|---|---|---|---|---|---|---|
| Out | Game | Platform | Type | Open | Meta | Steam | Meta | Hype |
The row is three claims, not nine columns: what critics scored it, what players made of it, and what was signalled before release. Each value carries the sample it came from underneath, and that sample turns red when it is too thin to trust — under 8 critics, or under 30 player ratings.
Because a 92 from twelve critics is weaker evidence than a 94 from a hundred and ninety. Each score is pulled toward the corpus mean (71.42) in proportion to how little supports it — a 25-review prior. The number displayed is always the real score; only the ordering is weighted, and hovering a score shows what it contributed. Checked against two Game Awards slates: this puts six of six 2025 nominees and five of six 2024 nominees in the top ten, against four and five when sorting by raw score. Score remains available and is unweighted.
They are one dial, not four alternatives: each nests strictly inside the last. Every hyped release is notable, because hype is one of the three things that make a release notable in the first place, and the counts on the buttons show that containment for whatever period is selected. Measured across the whole archive, 769 of 769 hyped releases are notable — so Hyped can only ever narrow the list, never reveal something Notable withholds. Notable fires on any of three independent signals: a Wikipedia article, IGDB hype of 10 or more, or at least 20 critic reviews. Three, because each misses what the others catch — hype scores a beloved back-catalogue port at zero, and Wikipedia lags on very new releases.
Metacritic's player sample runs a median of 128 ratings against Steam's 2,499 — twenty times smaller — and on a 0–10 scale a coordinated campaign drags it to the floor. The old ranking was almost entirely sports franchises sitting at 3 and 8 out of 100: that is review bombing, not reception. On Steam the same list reads Overwatch at 31% of 414,000 reviews, Monster Hunter Wilds at 49% of 325,000 and Tekken 8 at 56% of 71,000 — real player revolts against well-reviewed games. Because a thumbs-up ratio is not an averaged score the two are never subtracted: the corpus is fitted once (critic ≈ 0.38 × Steam% + 44 across 4,084 games) and the ranking is the residual. Games with fewer than 200 Steam reviews are excluded rather than ranked, because falling back to the Metacritic figure would reintroduce the very measurement this replaces.
Because a family counts games rather than memberships, so a cross-generation release is counted once: PlayStation holds 5,819 games, not the 9,880 you would get by adding PS4 and PS5 together. Platform offers families as well as exact entries, since "PlayStation" rarely means one specific console. Switch and Switch 2 stay separate at 4,480 and 222 — pick Nintendo for either.
Under 40 critics, a score of 82 or better, and at least 500 Steam reviews running 85% positive or higher. It is the one cut that does not nest, which is why it sits apart from the dial: it asks the opposite question. A niche release cannot be judged on critic volume it will never have, so it is judged on players instead — a game with six critic reviews and 26,000 player reviews at 98% is well evidenced, just not by critics. It needs Steam evidence, so the cut is effectively PC-only whatever the platform says, and coverage grows with the Steam backfill, making the list a floor rather than the full set.