What we study
Five lenses on player culture
Player sentiment
How players feel, and how that shifts over a game’s life.
Player communication
What players say to developers, and to each other.
Community language
The words and references a community shares.
Community culture
The norms and rituals that shape a player base.
Audience behaviour
When players arrive, stay, leave and come back.
Our evidence
Why player feedback matters
A review is written voluntarily, in the player’s own words, at a specific moment in their relationship with a game — hour one, hour one hundred, or year three. One review is an opinion. Two hundred thousand, handled fairly, is evidence about how a hobby behaves.
Large-scale public reviews are our core evidence base. We read them alongside wider community discourse and other open signals.
Choosing questions
Questions people actually ask
We start from what a player or a developer would ask — “when does a game’s audience turn?”, “who writes the angriest reviews?” — not from whatever the data makes easy.
A question makes the list when there is enough public feedback to answer it fairly, and when a clear answer would help the people who build, publish, cover or play games.
How findings are checked
Every headline has to survive a stress test
Recount three ways
Without the biggest games, on full-length reviews only, and within each genre group.
Correct for chance
When many ideas are tested at once, coincidences appear. We correct for that and keep weak effects out.
Hunt the single game
We check whether one game is secretly carrying an effect across a whole genre.
Ship the caveat
Every number travels with its sample size and its limits. If a claim can’t carry its caveat, it isn’t published.
The findings that didn’t make it
Most interesting-looking patterns do not survive, and that is the method working. One tempting genre-wide “sensation” — a free-to-play collapse in month two — turned out to be more than 99% one game, and was never published as a genre finding. We keep a register of what didn’t hold up, and we would rather publish five solid findings than fifteen shaky ones.
Independence
Independent, and open about mistakes
CritField is independent and not affiliated with any storefront, publisher or platform. Our studies use public Steam data; Steam is a trademark of Valve Corporation, and the project is not affiliated with or endorsed by Valve.
If a number is wrong, we fix it in the open and note what changed. Corrections and data questions go through the contact page.
Scope and limits
What we will not claim
- We study how players write and what that language signals — we do not aggregate review scores or rate games.
- Our findings are correlational and corpus-specific; they describe association, not cause.
- We report on the data we hold. Nothing here speaks for “all of Steam” or “all players”.
Work with the studio
Developers and publishers curious how these patterns look in their own genre, and journalists who want a comment or a chart, are welcome to get in touch.