Which claims are tested, and which are judged
What the README claims, and what holds each claim. A test is run by pnpm check; a script is run against the deployed app.
| Claim | How it is held |
|---|---|
| Every mechanic has a sourced decision, none overdue | spec/ledger.test.ts |
| A steer in one browser shows in another within a second | spec/realtime.test.ts |
| Other bikes move smoothly however late snapshots arrive | spec/netview.test.ts, on arrivals recorded live |
| No one can steer another person's bike; a room holds eight | spec/lobby.test.ts |
| The leaderboard, qualifications and log survive a restart | spec/persistence.test.ts |
| Every page passes axe, and every link here opens | spec/a11y.test.ts, spec/readme.test.ts |
| A blow moves like one on every frame, not only its key poses | spec/motion.test.ts, and a contact sheet before it ships |
| A punch lands with weight | Judged by playing it and reading the contact sheet; the tests reach only "not wrong" |
| Every helmet goes with its leathers | spec/liveries.test.ts |
| It looks like the original | Judged frame beside frame against 137 screenshots, each in a commit |
| It handles like the original | Ridden, then pinned as numbers in spec/feel.test.ts |
| It holds up over a bad network | Ridden at 300 ms with 3% loss, and with 8 scripted people |
| It works from the keyboard alone and survives a mid-race resize | scripts/rough.ts on the deployed app |
| Two people who follow this page race each other and reach the leaderboard | scripts/marker.ts on the deployed app, two browsers without shared cookies, one per marking viewport |