New Steam releases outperforming their release-week cohort. Tap a row for the full note.
What am I looking at?
Every new Steam release is tracked for its first 30 days, with a review-count snapshot
roughly once a day. Each game is compared to its cohort β games released within a
few days of it, at a similar price β at days 3, 7, 14, 21 and 30. A game gets
flagged when, at day 7 or later, in a cohort of at least 15 games, it has a robust
z β₯ 3 or ranks in the cohort's top 3 by reviews. Flagged games get a short
LLM-written note (tap the row). Click any column header to sort; hover it for a definition.
Robust z
How far the game's review count sits above its cohort's typical game, measured on
log-scale reviews in MAD units (median absolute deviation β like standard deviations,
but not thrown off by one mega-hit in the cohort). 0 β typical, 2+ notable
(orange), 3+ exceptional
(red) and part of the flag rule. Absurdly large
values (30+) just mean the cohort's median game has ~0 reviews, so any traction at all
divides by almost nothing β common in the first days of a release week.
Rank
Position within the cohort by review count (1 = most reviewed), shown as rank/cohort size.
Day
Which score day the row reflects: the game's age (in days since release) at its most
recent scoring, one of 3/7/14/21/30.
+7d
Reviews gained over the trailing 7 days β blank until a game has a week of snapshots.
Tag fit
LLM judgment (1β5) of how well the game's own Steam tags/genres describe it.
1 = the tags badly misdescribe the game. Low fit can mean discovery is broken β or
that the game found an audience its tags don't capture.
AI
π€ means the developer declared AI-generated content on the game's store page β
Steam requires developers to disclose it at submission. Hover the icon (or open the
row) for the developer's exact wording. Blank = checked, no disclosure; a small dot
means the store page hasn't been checked yet.
Core loop
One-sentence LLM summary of what the player actually does.