One Year of YiBoard: A Look Back
A year in, YiBoard went from one game to a set of parts that hold together. What shipped, which calls were wrong, and what we have decided not to build.
A game anniversary is usually the moment a studio posts a montage of its best moments. Let us run it the other way round: the wrong calls first, then what actually shipped. Over this year YiBoard moved from "can we get Gomoku running at all" to "how long does a new game take", and those two questions sit on opposite sides of an architecture. The first question is about one board. The second is about everything underneath it, which is where most of the work went.
What got built in year one
- The referee layer: legality and win detection written once for every game, so a new board does not mean a rewrite
- An alpha-beta engine running in the browser with a 500 ms search budget, which keeps server compute out of the picture
- Grades and ratings: everyone starts at 1200, Sixth Grade, settled one game at a time
- Five languages shipped together from day one rather than bolted on later
- Spectator mode, shareable match links, and full keyboard play
Looking back, the part that mattered was never Gomoku itself. It was the referee and ranking layers underneath, which carry over. Adding another game now costs an order of magnitude less work than it did a year ago, and that is the single number we watch when deciding what to build next.
Calls we got wrong
- We treated mobile as a smaller layout. Touch accuracy on stone placement was its own problem and forced a rebuild of move confirmation
- We underestimated translation upkeep. One changed interface string touches five languages, and we are still catching up
- The starting difficulty was too high for beginners, and we lost people in the first weeks before adjusting it
What people actually use
Playing without an account is the top path, and we guessed that correctly from the start. Sharing a match link came in second, which surprised us since it began as a debugging feature. Spectator mode sees less traffic than expected, but the people who use it stay for a long time, so it stays.
What we have decided not to build
- No ads and no paywall. Monetisation pressure on a board game pulls the rules design off centre
- No games added for the sake of a bigger list. Each one has to reuse the referee and ranking layers, or it waits
- No forced signup. Carrying a record across devices is worth an account; playing a game is not
- No spectator chat. If we cannot moderate it, we do not open it
The concrete roadmap items live in /blog/go-roadmap, most of the changes came out of /blog/iterating-on-user-feedback, and the measured numbers from the performance rounds are in /blog/performance-work. Background on the team and the project is on the about page.
FAQ
How long has YiBoard been running?
A year from the first commit. Most of that year went into making the referee and ranking layers reusable rather than polishing one game into perfection.
What changed the most this year?
The cost of adding a game. The first version of Gomoku took a long stretch of work; a new board now mostly means writing its rules, because the structure underneath already exists.
Which games come next?
Xiangqi and Go are on the roadmap, and /blog/go-roadmap has the order. We would rather not ship one just to make the list longer.
Is there an anniversary event?
No rating reset and no limited cosmetics. An anniversary should be invisible to the person who just wants to play.
Related
- From 1200 to 1800: A Rank-Up Roadmap
2026-10-13
- Mobile Game Optimization: A Board Built for One Hand
2026-10-20
- Why We Are Open Source
2026-10-29