The same opportunity scan a Claude session runs (screen-flip-niches.mjs) — identical gates and numbers, one table per niche (Band · Churn), each ranked by a letter Grade so the best flips sort to the top. Falling handling is per-strategy — band/churn skip fallers, but anything you hold, ask about, or watchlist still shows with price-to-clear guidance. On the deployed site this is a published snapshot (Refresh re-reads the committed screen.json); on the local dev server Refresh runs a real scan and regenerates the file live. Click any item for its live Trends.
Recent movement (last 2h, 5m steps)
Why this trend?
Price history (up to 3 months, 6h steps — pick a window, drag to pan, wheel to zoom)
Forward forecast (next 24h projection off the daily rhythm — provisional, n≈0)
Timing & seasonality
Average price by hour (your local time)
Trade volume by hour
Read the daily rhythm
Pick an item and analyse to get suggested buy times and prices. Hourly data starts at ~15 days (the API’s cap) and grows past a month as you keep it starred and refresh over the coming weeks. You can also click any item name in the Finder.
Item
Buy
Sell
Margin
ROI
Turn
Profit/hr
Grade
Items here are quarantined from the merch views (farming inputs, loot, personal-use). Their trades still stay in the fill log for the audit — this only hides them from positions/screens. Edits push to ignored-items.json when a GitHub token is set; the pipeline applies the filter on its next sync.
Item
Reason
GitHub sync — log & sync from mobile (no desktop file access)
Log a trade…
Leave “When” blank for “now”. Backdated trades must carry the real trade time — FIFO matching depends on it. Entries write to your linked fills log (the single source of truth); nothing is stored browser-only.
Open positions
Item
Qty
Avg buy
Sell now
Unrealised
Closed flips
P&L by
Item
Qty
Avg buy
Avg sold
Tax
Realised
Diagnostics — newest first, persists across launches (last 200 events). info · warn · error: your actions (tab switches, watchlist, trades, refreshes) alongside system events from market, guide and storage. Successful loads are logged too, so you can confirm a build and a data source are live.