Why We Deleted Most of Our Catalogue
This site used to publish close to a thousand games. It now publishes roughly a hundred and sixty. Nothing broke, nobody sent a takedown notice, and the games we removed are still perfectly playable elsewhere. We deleted them because the pages carrying them were not worth having.
That decision was not entirely voluntary, and pretending otherwise would be silly, so here is the whole sequence.
How the site got that big
Browser game distributors publish a feed. You agree to their terms, you get a licence to embed several thousand titles, and along with each game comes a title, a thumbnail, a category, a control list and a description. Building a site from that feed is close to free. One import script and you have a catalogue.
We ran that script. Every game in the feed got a page, and every page carried the description the feed supplied. It looked like a real site. It had category listings, a search box, thumbnails in neat rows, and enough pages to make the sitemap feel substantial.
The problem is that the description supplied with a game is supplied to everybody. The same sentences, in the same order, sit on every other site carrying that title, which was a lot of sites. Our pages were not similar to those pages. They were identical, minus the header.
The review that made us look at it
We applied to run advertising, and the application came back rejected with two reasons attached: replicated content and low value content.
Both were accurate. We had spent no time at all deciding what to publish, and no time at all describing it. What we had built was a mirror with a logo on top. The temptation at that point is to argue with the assessment, tweak the templates, add a paragraph of generated text to each page and reapply. We did consider it. It would have been faster.
What we actually did
We cut the published catalogue down to games somebody here had played, which turned out to be a much shorter list than we expected. Roughly a hundred and sixty titles survived. Everything else stopped being exported.
Then we wrote a guide for each survivor from scratch. Not a rewrite of the distributor’s copy — a fresh piece describing what the game is, how the opening minutes go, what the controls actually are, and two or three things we only learned by losing. That is the part that took months rather than days, and it is the reason the catalogue is capped where it is.
Five things the exercise taught us
Most of a big catalogue is one game repeated. Once you sit down with them, a feed of a thousand titles reveals maybe eighty distinct designs wearing different art. The moment we started asking “what does this do that something already on the shelf does not”, the list collapsed on its own.
A lot of games do not work. Not a trivial number, either. Dead menu buttons, a second level that never loads, a touch layout that needs a key press. None of this is visible from a thumbnail and a description, which is precisely why nobody publishing a feed at scale ever finds out.
Writing about a game forces you to have an opinion. It is impossible to write three hundred honest words about a puzzle without noticing whether it is any good. The write-up requirement turned out to be a quality filter as much as an originality one, because several games got cut at the point where somebody tried to describe them and had nothing to say.
The useful thing is a tip, not a summary. A description tells you a game exists. Knowing that you should count the visible units of a colour in Water Sort before committing, or that the tyres in Arcade GP tell you to pit long before the lap times collapse, is the thing that is worth a page.
Traffic follows the small list, eventually. We lost pages, and we lost the traffic attached to them. What came back was slower and stickier, because a page that answers the question somebody actually had is a page they return to.
What we do now instead
Every published game has an original write-up, and a build check refuses to ship a game without one. There are length limits so nobody can pad a guide or leave a stub, and a similarity check that fails the build if the same long run of words shows up across three or more guides. The full policy is on the about page.
We also started making our own games, which is the other half of the answer. Rush Sudoku, Word Circuit, Link Lab, Mini Clue Grid, Glow Grid and Pixel Trace are written here, run entirely in the browser, and carry no advertising inside the game itself. They are the only pages on the site where we control the whole thing.
Would we recommend it
If you run a site built on a syndicated feed: the rejection is not a formatting problem and you cannot template your way out of it. The question being asked is whether a person would rather land on your page than on the twelve identical ones next to it, and there is exactly one way to make the answer yes.
It costs more than the shortcut. It is also the only version of this that was ever going to work.