Market microstructure, made visible
Resting liquidity aggregated from nine venues and drawn where it sits. Every large executed trade on a live tape, with sound. Forced liquidations on their own strip. And a recorded history of all of it that no venue sells back.
Market structure is where the size sits.
Price gravitates toward large resting liquidity. Seeing that liquidity — across venues, at the price levels that matter for the symbol's scale — is the whole point.
Resting orders are claims. Executed prints are facts.
Walls get pulled; spoofing is real. So the two layers are never blended: what is resting and what has actually traded stay visually distinct on every surface.
Notional-summed, price-binned depth drawn as bars beside price — grouped relative to the symbol, so a BTC book groups at $100 to $10,000 and a DOGE book at fractions of a cent. A liquidity heatmap behind price, volume profile, session VWAP, and an estimated liquidation-cluster map, each toggled independently.
Spot and perpetual books from every connected venue, merged into one ladder with the basis between them shown rather than hidden. The largest resting walls are listed with their distance from mid — the levels a chart reader would otherwise have to hunt for.
Every trade above a per-symbol notional threshold, sized and coloured by side and venue, with a scheduled WebAudio note per print — compressed and paced so a burst of prints reads as a burst, not a machine gun. Forced liquidations are kept on their own strip, never mixed into the tape.
A rolling 24-hour table of executed flow per market — exchange by symbol — built from the collector's own recorded prints. Delta is shown as a share of that market's own volume, so it ranks one-sidedness rather than size.
The trading-hub clock — Asia, Europe, United States, after-hours — with each hub's share of volume measured from the last 30 days of recorded flow rather than quoted from folklore. The study's own figure sits beside each live one so drift is visible.
One failure-isolated adapter per feed. A venue going quiet costs its own data, never anyone else's. Fifty-two symbols are recorded around the clock, watched or not — book history cannot be backfilled.
One asyncio Python process. Venue adapters feed a delta-maintained book per market; aggregation bins the merged depth and broadcasts it over a WebSocket once a second, with trades streamed as they print.
SQLite in WAL mode: binned depth snapshots, large prints, per-minute flow, open-interest observations, forced liquidations. Per-table retention — and liquidation prints are never pruned, because that history has no free source anywhere.
React and TypeScript, built with Vite, served as static files by the collector itself. No server rendering, no Node in production, no API keys — only public market-data streams.
Market Lens is a visualization and measurement instrument. Markets — crypto especially — are volatile, and nothing shown by this tool is a recommendation to buy or sell anything. Every layer has limits, and the in-app documentation is explicit about them: depth is what venues choose to display, liquidation maps are estimates, and a wall is a claim until it trades.