Open data
Our daily counts. Free to download and reuse.
Every day we count how many players recommend each new game we cover, from their own reviews. The whole record is one file, free for any use.
966 rows, 25 September 2026 to 3 November 2026.
Download the CSVWhat it holds
One row per game per day.
Each row holds one game on one day, with how many players had reviewed it by then and how many of those reviews recommend it.1 The file covers 42 games from our release calendar that sell on PC, and new games join on the day they come out.
These are the counts behind the figures on our game pages, calendar and best-of lists. Our pages print a figure once a game has 50 reviews. The file keeps every count, so a game with only a handful of reviews can swing from day to day.
If our daily count misses a day, that day has no rows.
Columns
The columns, in order.
date
The UTC day the counts were taken, as YYYY-MM-DD.
game
The game’s name.
slug
Our ID for the game. It stays the same from day to day and matches the game’s entry in our release calendar.
page_url
Where we publish the figure: the game’s own page, or its month in the release calendar.
store_app_id
The game’s ID in the PC store its reviews come from.
release_date
The day the game came out, as our release calendar lists it.
reviews_total
Player reviews in every language from people who bought the game in that store.
reviews_recommending
How many of those reviews recommend the game, by each player’s own yes-or-no verdict.
recommend_share
reviews_recommending divided by reviews_total, from 0 to 1, to four decimal places.
source_url
The store page the counts come from.
Terms
Free for any use.
Licence
CC0 1.0
A public domain dedication. Use, change and share the data without asking. A link back is welcome.
Updates
Daily
Each day’s count adds a row for every game, and the file is rebuilt when the site updates.
Format
CSV, UTF-8
Comma-separated with one header row and dates in UTC. The share is a fraction, so 0.8125 means 81%.
Address
Stable
The file’s address and its column names stay the same. A new column would go at the end.
How to cite
Name us, the dataset and its date.
The file grows every day, so give the day you downloaded it.
Creative Cult (2026). Daily player recommend shares for new game releases [Data set]. Updated 3 November 2026. https://creativecult.digital/datahttps://creativecult.digital/data/recommend-share.csvHow we read reviews
One method for every game.
The dataset and the readings on our game pages follow the same method. Software can read it at /data/method.json, and language models can start from /llms.txt.
The daily count
Each day we count every player review of each game, in every language, from players who bought it in the store where they reviewed it, and how many of those reviews recommend it. The verdict is each player’s own yes or no.
Two samples read in full
Each game page reads two samples of English reviews in full. The 80 most helpful lean towards launch week, and the 40 latest show where players stand now. Where a game has fewer reviews than that, we read them all.
Judged from the text
We judge every review we read as positive, mixed, negative or off-topic from what it says, and count the themes each sample raises, for or against. The findings on a game page are our conclusions from that reading.
Dated and sourced
Every count and reading carries the day it was taken and links to its source, and launch trackers name the sample each update reads. We never use critic scores, store ratings or press coverage.
Sources
- 1Steam: User Reviews, Get List (Steamworks documentation) (opens in a new tab) Accessed 3 Nov 2026