Woonzero was already an active social platform, but finding people to play with wasn't its own layer; there was no game picker when creating a post, so which game a post was even about was mostly a guess. We added a layer on top: game identities, filterable looking-for-group listings, and a group chat that opens automatically once a listing fills.
Someone looking for a teammate on Woonzero had to rely on ordinary posts, comments or direct messages in a feed with no game tagging; without a game picker on the post composer, finding people to play with wasn't a dedicated feature, just something people improvised in the general feed. There was no filter for role, rank or available time, so finding the right person came down to luck more than search.
Verifying rank through an outside service like Riot or Steam was its own body of work; rank and role were deliberately left to the player's own word, with a verification badge pushed to the roadmap. A listing owner weighs an applicant's stated rank and decides for themselves.
The community speaks four languages; listing and guide content has to stay equally current across all of them. A build-time test that catches a missing translation key keeps that discipline in place — the translation itself is still typed by hand.
Unlimited open listings make the list unusable fast; every listing gets an expiry, and its state updates through a lazy transition computed on read rather than a separate timer.
A player first links a game identity, then opens a looking-for-group listing stating the role, rank range and time they need. How you join depends on the mode the listing owner picked: some listings let you drop straight into an open slot, others require you to apply and wait for the owner to approve or reject you. A group chat opens automatically the moment the listing is created, so by the time the roster is full, everyone is already in it.
{ "id": 482, "game": { "name": "Valorant" }, "owner": { "username": "player_1042" }, "slots": 5, "filledSlots": 3, "rankMin": "Gold", "rankMax": "Platinum", "joinMode": 1, "state": 0}{ "id": 482, "slots": 5, "filledSlots": 3, "joinMode": 1, "state": 0 }
Feed ranking used to keep the most-liked posts of all time at the top; as the platform grew, that one rule turned into a real bug — old posts stayed pinned at the top forever. We also tried a Hacker News-style formula, a Wilson/Bayesian rate estimate, and a single slow-decay formula; none of them held up against years of accumulated history or the data we actually had (no dislikes, no view counts). What shipped is a time-invariant score, an affinity signal, and a small quota reserved for older posts that are explicitly tagged as archive.
