Press room · Issue 01
Steam Reviews 202K — press kit
One sentence: CritField is an independent studio that reads player feedback at scale; our first study maps 202,000 Steam reviews to show where games lose their players — and is careful about what that can and cannot mean.
Published 2026 · Last updated 2026-07-23
The study in 60 seconds
We took an equal 50,500 reviews from games at each of four stages of life — the first month, months two to six, six months to two years, and older than two years — and compared the tone. Games get a honeymoon (17.7% negative in the first 28 days) and then a harder month (26.9% in days 29–180). The single toughest moment is the first two hours, the refund window, where 37.6% of reviews are negative — double the rate afterwards. Criticism also travels further: a negative review is about 3.6× as likely to collect five or more "helpful" votes. And past 200 hours, everything depends on genre: MMO and MOBA veterans are the fiercest critics (38–42%), while simulation veterans are the mildest (10%). All findings are correlational, specific to this sample, and not causal.
Key facts
- 202,000Steam reviews analysed
- 3,111games
- 15genre groups
- 4 × 50,500reviews per life-stage (balanced)
- 29review languages in the corpus (Jan 2025 – May 2026)
- Non-causalcorrelational, corpus-specific
Media angles (validated findings only)
37.6% of reviews written inside the Steam refund window (first two hours) are negative, versus 18.3% afterwards.
N = 36,803 refund-window reviews. Correlational; describes who writes and when, not a cause.
Negativity rises from 17.7% in the first 28 days to 26.9% in days 29–180 — the peak of the life cycle.
50,500 reviews per stage. Different reviews of different games per stage, not the same players over time.
A negative review is about 3.6× as likely to collect five or more "helpful" votes as a positive one, keeping it on the store page longer.
Vote counts also reflect Steam's visibility ranking; we see the combined outcome.
Among players with 200+ hours, MMO/MOBA veterans are the harshest critics (38–42% negative) while simulation veterans are the mildest (10%).
N = 10,287 veteran reviews; a cluster property, not a rule. Groups under 100 veteran reviews excluded.
About 2% of reviews receive an official developer response, and it goes to criticism ×4.3 more often than to praise.
Prevalence across the corpus; not a measure of whether replies change minds.
We deliberately do not offer the "free-to-play month-two collapse" as an angle: in this sample that pattern is over 99% one game (Stormgate) and is a single-game case, not a genre finding.
Ready-to-copy citation
For a specific number, add the chapter, e.g. "CritField, Steam Reviews 202K study, chapter 01: Lifecycle."
Methodology in brief
- Balanced sample: equal reviews per life-stage, balanced across genre and review-volume, with a per-game cap so no single release can dominate.
- Every headline recounted three ways (without the biggest games; full-length reviews only; per genre group) and checked for single-game concentration.
- Multiple-comparison correction applied; weak effects kept out of the conclusions.
- Full details on the methodology chapter.
What the study cannot claim
- No causal claims — reviews show how players describe their experience, not what caused it.
- No same-player-over-time claims — this is a cross-section of different reviews.
- No generalisation to "all of Steam" — it is a sample built for fair comparison.
- Review language is not nationality; review-volume tier is not budget.
Downloadable charts
Each chart is generated from the study's data and is self-labelled (source, sample size, caveat, CritField mark) so a screenshot is safe to publish. Dark- and light-background versions, SVG and PNG.
Logo & use
The CritField wordmark may be used to credit the studio in coverage. Reuse and attribution terms are on the use & credit page. In short: quote and screenshot freely with credit and a link; don't republish the study wholesale.
Press contact
Write to us for a quote, an interview, a figure checked before you publish, or a chart in a format your CMS will take. We answer questions about the data, and we would rather spend ten minutes on the phone than see a number travel without its caveat.
Deeper questions about sampling, method or the data behind a figure can go straight to [email protected].