Dark Pool Pulse — Off-Exchange Volume Scanner 📊

Changelog

• 2026-07-29 — Reliability fix: runs launched via API/integrations with unset optional fields sent as null could fail outright instead of using the documented defaults. Fixed, covered by tests.

Scan any US stock tickers and find out when their off-exchange ("dark pool") trading volume just spiked — and which venues are behind it.

Why dark pool volume matters

Most retail trading happens on public exchanges (Nasdaq, NYSE), where every order is visible in real time. Big institutional trades often happen instead on dark pools — private venues (ATS, "Alternative Trading Systems") where large orders can be filled without moving the visible price first.

Individual dark pool orders stay private, but by law every ATS must report its aggregate weekly volume per symbol to FINRA. That data is public and free — but it's raw, delayed, and split across dozens of venues, so nobody looks at it directly. This actor turns it into a signal: give it a ticker, and it tells you whether this week's off-exchange volume is unusually high compared to that ticker's own recent normal — a sign that a large player may be moving size quietly before it shows up in the price.

What you get

Two kinds of rows in one dataset:

ticker rows — one per flagged symbol:

Field	Description
week_start_date	The most recent week FINRA has published data for
total_shares	Total off-exchange shares traded across all dark pool venues that week
total_notional	Same, in dollars
avg_trailing_shares	The ticker's own average weekly off-exchange volume over prior weeks
spike_ratio	This week's volume divided by the trailing average — the headline number
num_venues	How many distinct dark pools reported volume for this ticker that week

venue rows — the top individual dark pools behind that week's volume for the same ticker, with shares, notional value, and their share of that ticker's total off-exchange volume.

Only tickers that clear both a minimum volume and a minimum spike ratio are returned. Zero flagged tickers = never billed.

How it works — read this

Data comes straight from FINRA's official OTC Transparency API (api.finra.org), the same body that regulates US equity markets. No scraping fragility, no login, no key — it's a public, government-mandated disclosure. Reporting is delayed (2 weeks for S&P 500 / Russell 1000 names, 4 weeks for everything else), so "latest week" means the most recent week FINRA has actually published, not the calendar-current one.

The spike ratio is computed per ticker against its own trailing history (configurable lookback window), not against other tickers — so a normally-quiet small-cap and a normally-busy mega-cap are each judged against their own baseline.

Use cases

• Smart money tracking — spot tickers where off-exchange volume just broke out of its normal range.
• Pre-news positioning — unusual dark pool activity sometimes precedes a public catalyst.
• Venue research — see which specific ATS venues are moving size on a given name.
• Model inputs — feed the spike ratio into your own screening or backtests.

Input

Field	Type	Default	Description
tickers	array	20 liquid, volatile large caps (AAPL, TSLA, NVDA, AMD, META...)	US stock/ETF symbols to scan
weeks_lookback	integer	8	How many prior weeks to average as the baseline
min_spike_ratio	number	1.5	Minimum multiple of the trailing average to flag a ticker
min_weekly_shares	integer	100000	Minimum latest-week off-exchange share volume to consider (filters illiquid noise)
top_venues_per_ticker	integer	5	How many top dark pool venues to list per flagged ticker
max_results	integer	100	Cap on ticker rows returned
request_timeout_secs	integer	30	HTTP request timeout

Output example

{
  "row_type": "ticker",
  "ticker": "NVDA",
  "week_start_date": "2026-07-06",
  "total_shares": 87804579,
  "total_notional": 17642247746.85,
  "avg_trailing_shares": 81326827.14,
  "spike_ratio": 1.08,
  "num_venues": 30
}

Notes

• This is not investment advice — it surfaces a public data signal, not a trade recommendation.
• A ticker not appearing in the results just means its off-exchange volume didn't clear your thresholds that run, not that nothing happened on it.