Methodversion v0.7
How every number is made
Steam does not publish sales. Every figure here is an estimate built from public store signals: review counts, prices, release dates, tags, followers and Steam's wishlist ranking. This page says how each one is made, how much data it needs before we show it, and how far to trust it.
Figures on the site were computed from data of 30 Sep 2026. Methodology version v0.7.
Price
launches 2024 Q4 to 2025 Q3Every genre and niche has one headline price: the median estimated first-year revenue of its indie paid launches over the last four complete quarters.
price = median of (estimated revenue in each launch's first 12 months)
- Indie means a list price under $40 and not released by one of 30 named major publishers. Free games have no price, so they never enter it.
- Complete quarter means every game launched in it has had 12 full months of reviews.
- The change beside a price compares it with the four quarters before. It needs 10 or more launches in that earlier year.
- Minimum sample: shown with 30 or more launches; 10 to 29 is labelled thin; under 10 it is not shown.
- A median, not an average: a few huge hits would drag an average far above what a typical launch earns.
Stability: across genres the price ranking re-ran at 0.91 (pairs 0.83). See how stability is measured.
Revenue from reviews
Sales are private; review counts are public and track sales closely. We use the count Steam shows on the store page, which only counts reviews from Steam purchases.
revenue = reviews x sales per review x list price x 0.675
| Launch year | Sales per review |
|---|---|
| 2020 and later | 30 |
| 2018 to 2019 | 40 |
| 2015 to 2017 | 50 |
| 2014 and earlier | 70 |
- 0.675 is the share of list price a sale realises after discounts and regional prices.
- Games sold mostly in China earn less per sale: the estimate falls by up to 45%, in proportion to the share of Chinese-language reviews.
- Error band: published sales-per-review ratios run from 20 to 55 and realised share from 0.60 to 0.75, so one game's true figure can sit anywhere from about 0.6x to 2x our estimate. Medians over hundreds of launches are far steadier than any single game.
Launch dates
A game's launch is its first day on sale on Steam, including Early Access. For a game that has left Early Access, Steam's release date is the 1.0 date, which would make Palworld a 2026 launch.
We use Steam's original release date when it is earlier, or the first month with reviews when reviews start two or more months before the stated date. Ages, cohorts and the sales-per-review year all run from this launch.
Checked against 40 games with known launch dates, the rule is right within 31 days for 38 of them. The store date alone was right for 18.
What counts as a genre
A genre is a single Steam tag that describes what the game is. 167 tags qualify. The boards rank genres; inside each genre its best tag pairs (niches such as Roguelike + Deckbuilding) are ranked the same way.
- Tags about a game's theme (Space), mood (Cozy), features (Crafting), players (Co-op) or hardware (VR) are not ranked as genres. They still appear inside pairs.
- Presentation and umbrella tags (Pixel Graphics, 2D, Action, Casual) only appear inside pairs.
- A game belongs to a tag when the tag is among its 10 most-voted tags.
- Adult-only games are left out of every board.
The boards
data of 30 Sep 2026Four boards rank genres. Each value is shown with its range (the interval we are 90% sure it sits in) and the number of games behind it.
- Most anticipated
- Total wishlist heat: the genre's share of all wishlist-ranked upcoming games, weighted by rank, shown with the number of upcoming games in the genre.
- Minimum: 10 ranked upcoming games (pairs 5).
- Stability: 0.87 across genres, 0.68 across pairs.
- Predicts launches? Not testable yet: wishlist rank history started on 30 Sep 2026.
- Ranked now: 133 genres.
- Launch success
- How many indie paid launches cleared an estimated $100k, against how many would have for the same prices and launch dates anywhere on Steam. Above 1.0 means better than expected.
- Minimum: 100 launches (pairs 30).
- Stability: 0.89 across genres, 0.63 across pairs.
- Predicts launches? Validated. Tested from 2024 data, it ranked genres close to the order of their next year's launch results.
- Ranked now: 115 genres.
- Fewer new releases
- Fewer launches relative to players. Historically this did not predict better launches.
- Players gaining (as in Rising demand) minus the growth in new releases, compared with Steam as a whole. High means releases grew more slowly than the audience.
- Minimum: 20 releases in each 12-month period, plus the Rising demand floors.
- Stability: 0.85 across genres, 0.63 across pairs.
- Predicts launches? No. Genres whose releases grew had slightly better launches the next year, so fewer releases is not a sign of room.
- Ranked now: 105 genres.
- Rising demand
- Gaining players faster than same-age games. Descriptive, not a launch signal.
- How much faster the genre's games gained reviews over the last three months than a year earlier, compared with games of the same age elsewhere.
- Minimum: 1,000 reviews in each window and 30 games (pairs 15).
- Stability: 0.65 across genres, 0.40 across pairs.
- Predicts launches? No. It carries over from one year to the next, but it did not predict launch results. Pair readings are indicative only.
- Ranked now: 105 genres.
- Gaining fast
- Collecting since 30 Sep, first readings 13 Oct.
- Week-over-week follower growth of a genre's upcoming games. It needs 14 days of daily follower counts, so nothing is shown or estimated before then.
- Minimum: 10 games with counts on both days.
How stability is measured
We split the games into two random halves, rank genres on each half separately, and compare the two rankings, 20 times over. The figure is the agreement expected between our ranking and a re-run on equally good fresh data: 1.0 is identical, 0 is no agreement. Measured on 1 Oct 2026.
Launch success currently reads indie paid launches from 31 Mar 2024 to 1 Apr 2026: games 6 to 30 months old, long enough for their launch to have played out.
Noise rules
- Minimum samples
- Each board has a floor (above). Rows below it are kept but never shown.
- Shrinkage
- Small genres are pulled toward the Steam-wide value, so a handful of lucky games cannot top a board.
- Ranges
- Every board value carries a 90% range from resampling the games, or from the hit count for Launch success.
- Flags
- One hit carrying the genre, big publishers above 85% of it, Chinese-market games above 35%, or mostly free-to-play. Where tests showed a flag makes the reading less reliable, the genre is ranked below unflagged ones.
- Near duplicates
- Two genres sharing a third or more of their games (Roguelike and Roguelite) are one market: only the higher one is ranked.
Release calendar
checked 1 Oct 2026The calendar lists only games with a real date close to release, and re-checks every one against the live store before it is shown.
- Near term: a day or month date starting within 90 days. 30 shown now.
- Later: a day, month or quarter date within a year. 30 shown now.
- Announced, no date: TBA, year only, more than a year out, or a date that has passed. Listed separately, never on the calendar.
- Order: Steam's wishlist rank combined with the follower count. Wishlist counts are not public and we don't estimate them; rank and followers are shown with the date each was read.
- Check: still on sale soon, still a game, date unchanged or moved within its band, name and developer present. Up to 50 per band are checked and the first 30 that pass are shown.
Validation
GMD Verdicts, one call per niche that weighs these signals together, are in development. Validation is in progress: the combined model is being tested against past launches, and nothing from it is published until that test is final.
Until then every board is shown as what it is: a measurement with its sample, range and date, and a plain note on whether it has predicted anything.
Join the early list to hear when Verdicts open.
Limits
- Estimates, not sales. Every revenue figure comes from review counts. Treat single games as rough and medians as the useful number.
- Demand sees part of Steam. Rising demand uses monthly review histories for games launched since March 2021: about 38% of all reviews and roughly half of recent review flow. Genres carried by older evergreen games read low.
- Tags are today's. Players change tags over time and we only see tag history from the day collection began, so older launches are grouped by their current tags.
- The past is not the future. Boards describe the last one to two years. Only Launch success has shown it carries forward.