What this instrument measures
Radar is a window, not a market. It asks mempool.space for the newest confirmed blocks, then reads every transaction in those blocks and keeps the ones whose output sum clears the filter. That sum includes change returning to the spender. A 120 BTC row can be a 5 BTC payment plus 115 BTC change. Quote the explorer’s input/output counts before you quote the BTC column.
Coverage is complete for that window and incomplete for history. We do not scan the mempool tip, we do not keep a year of blocks, and we do not cluster unlabeled addresses into an entity. An empty ≥500 BTC log after eight blocks is a real absence in that slice of chain — not a feed outage. The depth field plots a subset of the largest rows (newest toward the right). Click a blip or use Brief → to run the same classifier the tests use.
The scanned-heights strip lists every block in the current window. Each tick is a height you can open on mempool.space. If a height is missing, the scan is still in progress or that block failed to fetch — the status line will say so. Eight blocks is an editorial window (~80 minutes on average). It is not an archive and it is not “the market today.”
Mini glossary for a row: BTC is output sum (change included). sat/vB is inclusion priority, not importance. in / out on a label is which side of the tx matched the directory. The vertical ping next to the amount is log-scaled size in this window — a visual, not a sentiment bar. Depth-field blips are a subset of the largest rows; the log below is the full filter result.
Refuse
unlabeled is the normal case. It means no exact directory hit — not “unknown to the world,” and not a license to invent an entity from size. USD is CoinGecko spot times output sum. If the price API fails, the row shows a dash. Neither figure is a trading signal.
Primary API:
mempool.space REST
(/api/blocks, /v1/block/{hash}/summary).
UTXO model:
Bitcoin Developer Guide — Transactions.
Longer treatment:
reading large transactions,
mempool vs confirmed.