Troubleshooting

Start here when something looks wrong. Most of these are documented limits rather than defects — and a few things that look broken are the app refusing to guess.

Browser and screen

The app targets desktop Chromium browsers (Chrome, Edge, Brave). It needs a window at least about 1280 px wide: the trading layout is a multi-pane desk, and it does not reflow into a phone.

Safari is best-effort. It is not a supported target yet: a WebKit gate (storage close/transaction behaviour, permission prompts, cache and service-worker handling, performance) is still open, so expect rough edges. If your work depends on Safari, test it on your own machine before relying on it.

If text looks unreadably small, raise the browser zoom; the interface enforces a minimum font size and does not shrink below it to fit panes.

Import

Nothing imported after a file drop: check the validation report first — malformed rows are reported and skipped, and an unrecognized file fails with a message and leaves your library untouched.
Suspiciously short history: look at the coverage section. Many free sources are incomplete outside peak hours; the report shows the time-of-day coverage so a gap is visible instead of invisible.
Fails near the end: this is usually browser storage. Free space, close other tabs, retry; the app reports quota failures instead of writing partial data.
A tick file imported as candles: tick archives must keep bid/ask; without both prices, real spread is impossible and the importer says so.

Replay and charts

Prices differ from your broker: the fills come from your dataset plus the spread setting you chose. Real recorded spread requires a tick dataset; otherwise spread is a number you set.
A same-bar stop and target both touched: with candle data the engine applies a documented convention; with tick data the actual sequence decides. Same-bar collisions are data-dependent, not a coin flip in the app.
No drawings or news markers on Renko/Range/Line Break: expected. Those panes have synthetic horizontal coordinates, so time-anchored overlays are hidden rather than placed wrong.
An alert did not fire again: fired alerts stay fired until re-armed; the Object tree's Alerts section has re-arm and clear-fired controls. Up to 50 active alerts per test.
A second monitor or window is missing: browser dialogs are blocked by the browser; the browser's own pop-up permission is the fix.

Sync and account

Not syncing: the offline banner is the first clue — writes queue while disconnected and flush when the link returns. A rejected push keeps the local write.
Older version of a test appeared: sync is last-write-wins per record; a later write on another device legitimately wins.
Signed out unexpectedly: sessions are capped per account (three devices) and expire when idle. Sign in again and revoke sessions you no longer use.
Magic link did nothing: links are single-use and expire in 15 minutes; request a fresh one in the browser you intend to use.

Payment still pending

Confirm the network matches the invoice (TRC-20 vs BEP-20), the amount matches the quote, and the destination is this invoice's address. Then wait for the confirmations shown on the invoice — the invoice state is authoritative. If it is still wrong after the network has clearly confirmed, contact support with the transaction id.

What is not a bug

No profit guarantees, no signals, no live trading, no broker connectivity — this is a backtesting and replay tool.
Marketing pages that 404 when served by a raw static file server: clean paths need the preview script (npm run preview) or the real host. 404.html is generated for the real host, not for a local file server.
Historical data is not redistributed by us; you import it from your own source.

Data & Storage for quota and backup specifics.
Billing for the invoice lifecycle in detail.