Dark Pool Data: Where It Comes From and How to Read It
Nizar Assanie
Dark pool data is the public record of trades that happened on private, off-exchange venues. In the US, most dark pool data comes from FINRA, which publishes weekly volume for every dark pool two to four weeks late, plus daily files that many websites mislabel as "dark pool" signals.
A dark pool is an alternative trading system (ATS) that does not show its orders before a trade. Every dark pool trade still prints on the public tape within 10 seconds, but the tape does not say which dark pool handled it or who was buying.
This guide is for self-directed investors who see dark pool charts, "short volume" numbers or print alerts and want to know what they really measure. You will learn where each dataset comes from, how to read one stock's numbers end to end, and which readings are noise.
Key takeaways
FINRA publishes weekly dark pool volume for S&P 500 and Russell 1000 stocks about two weeks after the week ends, and about four weeks later for other stocks. The week of Oct 19, 2026 was published on Nov 9, 2026.
FINRA's daily short sale volume files are not short interest. Apple's off-exchange short volume ratio was 53.5% from Oct 19 to 30, 2026, while only 1.0% of its shares were sold short.
In our test of 29 large stocks in 2026, a spike in weekly dark pool volume did not predict the next week's return: +0.48% after spikes versus +0.45% for all weeks.
Dark pool data shows size, price and time. It never shows the buyer, the seller or which side started the trade.
What is dark pool data and where does it come from?
Dark pool data is any record of trades executed on an ATS that hides orders before the trade. It starts with one rule: a broker that executes a trade off an exchange must report it to a FINRA Trade Reporting Facility (TRF) within 10 seconds, under FINRA Rule 6380A.
A Trade Reporting Facility (TRF) is a FINRA-run system that collects trade reports for stocks traded away from exchanges and sends them to the consolidated tape, the public feed of all US stock trades. Dark pools, wholesalers and bank desks all report through TRFs.
From that single stream of reports, four kinds of public data are built:
- Trade prints: each off-exchange trade appears on the tape with price, size and time, tagged as a TRF trade. The tape does not name the venue.
- FINRA weekly ATS data: share and trade counts for each dark pool and each stock, published two to four weeks later.
- FINRA monthly block statistics: how much of each dark pool's volume came in trades of 10,000 shares or more.
- FINRA daily short sale volume: how many off-exchange shares were marked "short" each day, published by 6:00 pm ET the same day.
What does a single dark pool print show?
A dark pool print shows the stock, the price, the number of shares and the time. It does not show the buyer, the seller, the dark pool's name or whether the trade opened or closed a position. Any tool that labels a single print "bullish" is adding a guess that is not in the data.
Most off-exchange volume is not dark pool volume at all. Cboe reported that 50.6% of US stock volume traded off exchange in 2025, but only 18.7% of that went through ATSs. Wholesalers filling retail orders handled most of the rest, so "off-exchange" and "dark pool" are different numbers. Our guide on what a dark pool is and how dark pool trading works covers the venues themselves.
Where can you find dark pool data for free?
FINRA is the only free primary source of dark pool data in the US. Everything else, from charting add-ons to paid print feeds, repackages FINRA files or the consolidated tape. The table shows what each free source covers and how late it arrives.
| Dataset | What it shows | Delay | Best for | Weak spot |
|---|---|---|---|---|
| FINRA weekly ATS data | Shares and trades per dark pool, per stock | About 2 weeks (Tier 1), 4 weeks (Tier 2) | Seeing how much of a stock trades in dark pools | Too late to trade on |
| FINRA monthly block statistics | Block trades (10,000+ shares) per dark pool | About 1 month | Finding block-focused venues | No stock-level detail |
| FINRA daily short sale volume | Off-exchange shares marked short each day | Same day | Market-maker activity research | Widely misread as short selling |
| FINRA short interest | Shares actually held short | Twice a month, about 1 to 2 weeks later | Real bearish positioning | Snapshot, not a flow |
| Consolidated tape prints | Every off-exchange trade, with size and price | Real time via a broker or data vendor | Spotting single large prints | No venue, no side |
How delayed is FINRA dark pool data?
FINRA weekly dark pool data is about two weeks late for Tier 1 stocks and four weeks late for Tier 2 stocks. Tier 1 covers the S&P 500, the Russell 1000 and selected ETFs. In 2026, the week of Oct 19 to 23 for Tier 1 stocks and the week of Oct 5 to 9 for Tier 2 stocks were both first published on Nov 9.
What does dark pool data cost?
FINRA's weekly, monthly and daily files are free on its OTC Transparency site and through its public API. Real-time print feeds cost money because they come from consolidated tape data that vendors license. TruthSayer AI's Dark Pools AI alerts are free; the app's AI chat costs $10.99 a month or $71.88 a year after a 7-day trial.
Before you trust any dark pool chart, find the line that names its source. If it says "FINRA short volume", it is the daily short sale file, not dark pool volume and not short interest.
How do you read dark pool data for one stock? A worked example
The best way to read dark pool data is to turn raw volume into ratios and compare them with the stock's own history. Here is Nvidia (NVDA) for the week of Oct 19 to 23, 2026, the latest Tier 1 week FINRA had published on Nov 11, 2026.
- Pull the dark pool volume. FINRA's weekly ATS data shows 65,430,235 NVDA shares across 798,341 dark pool trades. That is 65,430,235 ÷ 798,341 = 82 shares per trade, so these are small slices, not blocks.
- Pull the other off-exchange volume. FINRA's non-ATS data shows 158,344,990 shares traded off exchange outside dark pools, mostly through wholesalers and bank desks.
- Work out the dark pool share. 65.43 million ÷ (65.43 million + 158.34 million) = 29.2% of NVDA's off-exchange volume ran through dark pools.
- Compare with a baseline. NVDA averaged 69,473,195 dark pool shares a week over the prior eight weeks (Aug 24 to Oct 16). 65.43 million ÷ 69.47 million = 0.94 times normal. No spike.
- Check short positioning the right way. FINRA short interest for Oct 30, 2026 was 279,579,505 shares. Nvidia's 10-Q cover page lists 24.1 billion shares, so 279.58 million ÷ 24.1 billion = 1.16% of shares sold short, with 2.8 days to cover.
- Note the timing. The week ended Oct 23 and FINRA published it on Nov 9, 17 days later. Anything you conclude describes late October, not today.
The verdict for NVDA that week: normal dark pool activity, low short interest and nothing that calls for action. That is the most common result, and it is useful, because it stops you from chasing a chart that only looks dramatic.
Why does the baseline matter so much?
Dark pools' share of a stock's volume shifts for structural reasons, so a single week means little without the stock's own history. Nvidia's dark pool share of off-exchange volume rose from 18.0% in January 2026 to 28.8% in July while its price moved sideways to up, as the chart shows.
Nvidia: dark pools' share of off-exchange volume, 2026
A reader who saw only July might have called it "heavy dark pool accumulation". The steadier explanation is that brokers routed more NVDA flow to ATSs. Compare each week with the prior eight weeks, and avoid baselines that include holiday weeks, when volume is unusually thin.
Is dark pool short volume the same as short selling?
No. FINRA's daily short sale volume counts off-exchange trades marked "short", and most of that marking comes from market makers selling to buyers before they own the shares. It is not the number of shares investors have bet against, which is short interest.
"Although some websites redistribute the Daily File and refer to the data as 'short interest,' it is not, in fact, the equivalent of reported short interest information."
— FINRA, Information Notice on short sale volume data (May 10, 2019)
Our data shows the gap clearly. From Oct 19 to 30, 2026, Apple's off-exchange short volume ratio was 53.5% and GameStop's was 55.8%, almost the same. Yet GameStop's short interest was 8.40% of its shares and Apple's was 1.0%. The short volume ratio could not tell a heavily shorted stock from a lightly shorted one.
Dark pool data trap: short volume ratio vs real short interest
Short volume ratio, Oct 19 to 30Short interest, % of shares outstanding
This is the contrarian point of this guide. The most popular "dark pool" number online, the short volume ratio, sits between 30% and 60% for almost every large stock, every day. FINRA's notice gives the reasons: the file leaves out offsetting buys that are not publicly reported, and it covers only off-exchange trades, not exchanges.
A post that says "60% of today's volume was short, so the stock will fall" is reading the daily short sale file. Check real short interest and days to cover before you draw any bearish conclusion.
Does weekly dark pool data predict stock returns? Our test
Weekly dark pool data did not predict next-week returns in our 2026 test, and a hint of an effect after publication was too weak to trade on. We tested it because many tools sell dark pool volume spikes as signals.
Method. We took 29 large US stocks (Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta, Tesla, JPMorgan and 21 others) and FINRA's weekly ATS volume from Nov 3, 2025 to Oct 19, 2026. A "spike" was a week with dark pool volume at least 1.5 times the prior eight-week average; a "drop" was 0.67 times or less. We measured each stock's return in two windows: the week right after the data week, and the week FINRA actually published it, three weeks later. We excluded Netflix because its 10-for-1 split in November 2025 distorts share counts.
Does weekly dark pool data predict returns?
Week right after the data weekWeek FINRA published the data
- Next week: after 122 spikes the average return was +0.48%, against +0.45% for all 1,247 stock-weeks. Only 49.2% of spike cases rose, versus 52.5% overall. No edge, even with data you could not have had in time.
- Publication week: after 121 spikes the average was +1.24%, versus +0.36% for all weeks, about 1.14 points better than the average of the 29 stocks in the same week.
- The catch: 28 of the 121 spike cases fell in January, when the eight-week baseline included holiday weeks. From February on, the edge shrank to +0.85 points with a t-statistic of 1.76, below the usual bar of 2 for a reliable result.
Many volume spikes also happen in earnings weeks, so part of what remains may be post-earnings announcement drift rather than anything about dark pools. One year and 29 stocks is a small sample. As we explain in backtested vs live-traded signals, a result like this needs a live, out-of-sample test before anyone trades it.
Which dark pool readings matter and which are noise?
A dark pool reading matters when it is large relative to the stock's normal volume, unusual against the stock's own history, and confirmed by price. Everything else is usually noise. Use this guide when you look at any dark pool number.
| Reading | Usually noise | Worth a closer look | Strong evidence |
|---|---|---|---|
| Single print size | Under 10,000 shares or under 10% of average daily volume | 10% to 50% of average daily volume | Above 100% of average daily volume |
| Weekly dark pool volume vs 8-week average | 0.7 to 1.3 times | 1.3 to 1.5 times | 1.5 times or more, outside holiday and earnings weeks |
| Short volume ratio | Any level from 30% to 60% | Not a useful reading on its own | Use short interest instead |
| Short interest, % of shares | Under 2% | 5% to 10% | Above 10%, or days to cover above 5 |
| Price after a large print | Back to the print price within a day | Holds above or below for 2 to 3 days | Breaks away on rising volume |
These thresholds are our rules of thumb for reading the data, not universal laws. A 10,000-share print is a block for a $10 stock but a routine trade for Nvidia, so always scale size to average daily volume.
How does dark pool data compare with other ways to track big money?
Dark pool data is one of six common ways retail investors try to follow institutions, and it is neither the fastest nor the most precise. Each tool answers a different question, so the right one depends on your holding period.
| Method | What you learn | Timing | Cost | Best for |
|---|---|---|---|---|
| FINRA weekly dark pool data | How much volume ran through dark pools | 2 to 4 weeks late | Free | Research and context |
| Real-time print feeds | Single large trades as they print | Seconds | Paid | Active traders |
| Form 13F filings | Fund holdings at quarter end | Up to 45 days after the quarter | Free | Long-term investors |
| Form 4 insider filings | Executives' own buys and sells | Within 2 business days | Free | Spotting insider conviction |
| Options flow | Large options orders | Real time | Mostly paid | Short-term traders |
| A live-traded signal service | A tested entry, stop and exit | Hours | Varies; TruthSayer signals are free | Investors without time to filter prints |
- Doing it yourself with FINRA data: free and transparent, but slow and easy to misread.
- Paid print feeds: fast, but they show thousands of prints a day with no side attached, so the filtering is up to you.
- Filings (13F, Form 4): show who owned what, but long after the trades.
- Signal services: save time, but only worth using if they publish live results with losses. Our checklist for how to judge a signal service covers what to ask.
Which dark pool data source is best for you?
The best dark pool data source depends on how long you hold a stock and how much time you have each week.
Best for long-term investors: FINRA weekly data plus 13F filings
If you hold for months, the two-week delay does not matter much. Use FINRA weekly data to see whether institutions are active in a stock, and 13F filings to see which funds own it.
Best for swing traders: real-time prints filtered by average daily volume
If you hold for days to weeks, only real-time prints arrive in time. Filter to prints above 50% of average daily volume and wait for price confirmation before acting.
Best for beginners: a service that shows its filter and its losses
If you are new to market structure, a filtered alert that names the print size, the entry, the stop and the exit teaches more than a raw feed. Our explainer on AI stock signals shows how to judge one.
Who should not use dark pool data
Dark pool data is not right for day traders who need second-by-second order flow, or for anyone who wants a single number that says "buy". It answers "how much traded where", not "what happens next".
What are the common mistakes when reading dark pool data?
The most common mistake is treating volume as direction. Every dark pool trade has a buyer and a seller, so volume alone cannot be bullish or bearish.
- Calling short volume "short selling". The short volume ratio sat between 32.8% and 55.8% for all eight stocks in our chart, regardless of real short interest.
- Using raw shares instead of ratios. NVDA's dark pool shares fell from 82.2 million in the first week of January to 65.4 million in the week of Oct 19, 2026, while its dark pool share of off-exchange volume rose from 18.3% to 29.2%.
- Ignoring the delay. FINRA weekly data is 2 to 4 weeks old when it appears.
- Forgetting the baseline. Holiday and earnings weeks create false spikes.
- Mixing up "off-exchange" and "dark pool". In 2025, only 18.7% of off-exchange volume traded on dark pools, according to Cboe.
The honest limit of dark pool data is that it describes where trades happened, not why. Use it as context alongside price, earnings dates and filings.
How should you check a dark pool data claim? The TruthSayer Dark Pool Data Audit
The TruthSayer Dark Pool Data Audit is a five-step check for any dark pool chart, alert or social media post. It takes about five minutes per stock.
- Name the dataset. Is it FINRA weekly ATS data, a single print, the daily short sale file or short interest? If the source is not named, stop.
- Check the date. Write down the trade date and the publication date. Weekly data is 2 to 4 weeks old.
- Turn shares into ratios. Divide a print by average daily volume, and weekly dark pool volume by the prior eight-week average.
- Replace short volume with short interest. Use short interest as a percentage of shares outstanding and days to cover.
- Wait for price. A large print matters only if price holds away from the print level over the next two to three days.
How does TruthSayer AI use dark pool data?
TruthSayer AI's Dark Pools AI uses real-time prints, not delayed weekly files. It scans market data for unusually large institutional transactions, filters for statistically significant activity, confirms price direction, and then trades the signal in TruthSayer's own funded account before the alert goes out.
A real example: on Aug 18, 2026, Dark Pools AI flagged three dark pool trades in Thomson Reuters (TRI) totaling 4,470,000 shares, 204% of its average daily volume. The model went long at $102.04 with a 15% trailing stop and exited on Aug 20 at $106.61.
- Position. 15 shares × $102.04 = $1,530.60 invested.
- Exit value. 15 × $106.61 = $1,599.15, a gross gain of $68.55.
- After costs. $68.55 − $9.98 commission = $58.57. $58.57 ÷ $1,530.60 = +3.83% in 2 days.
Not every large print works. On Aug 25, 2026, the model went long Sun Life Financial (SLF) after 5,700,000 shares printed, 792% of average daily volume, and was stopped out on Oct 1 for −3.48%. Both trades are in the public trade list.
As of Nov 11, 2026, the Dark Pools AI index shows a 33.62% return since inception against 29.44% for the S&P 500, a 67.3% win rate, a Sharpe ratio of 2.78 and a beta of 0.91. TruthSayer reports performance under the GIPS standards, with losing trades included.
The reading thresholds in this article are our rules of thumb, checked against the FINRA and SEC data analyzed for this guide. TruthSayer's own account adds fixed risk rules: 5% position size and a 3× ATR trailing stop. We picked the TRI and SLF trades because they are the most recent closed Dark Pools AI trades with prints above 200% of average daily volume, one win and one loss.
What should you do next?
Start with one stock you own and run the TruthSayer Dark Pool Data Audit on its latest FINRA week. If the numbers are normal, as they were for Nvidia, you have saved yourself from a false signal.
- Read what a dark pool is for how the venues work and how big they are.
- Read how to judge a signal service before paying for any dark pool alerts.
- Read backtested vs live-traded signals to see why our 2026 test is a starting point, not proof.
- See live alerts, entries and exits, losses included, on the Dark Pools AI page.
- Look up terms like ATS and short interest in our glossary.
The bottom line: Dark pool data is useful context and a poor crystal ball. FINRA's weekly files are free but two to four weeks late, and the popular short volume ratio is not short selling. Run the TruthSayer Dark Pool Data Audit before you act on any dark pool claim: name the dataset, date it, turn it into ratios, check real short interest and wait for price.
Frequently asked questions
Where can I get dark pool data for free?
FINRA's OTC Transparency site publishes free weekly dark pool volume by venue and by stock, plus monthly block statistics. Tier 1 data arrives about two weeks after the week ends; the Oct 19 to 23, 2026 week appeared on Nov 9. FINRA's daily short sale volume files are also free, but they measure something different.
Is dark pool data real time?
Individual off-exchange trades are real time: FINRA Rule 6380A requires reporting within 10 seconds, and they appear on the consolidated tape. The tape does not name the dark pool. FINRA's venue-level dark pool data is delayed about two weeks for Tier 1 stocks and four weeks for Tier 2 stocks.
Does a high short volume ratio mean a stock will fall?
No. The short volume ratio mostly reflects market makers filling buy orders. From Oct 19 to 30, 2026, GameStop's ratio was 55.8% and Apple's was 53.5%, yet GameStop's short interest was 8.40% of shares and Apple's 1.0%. FINRA states the file is not short interest.
What should I do after I spot a big dark pool print?
Divide the print by the stock's average daily volume, then watch price for two to three days. A print below 10% of average daily volume is usually routine. If price returns to the print level within a day, treat it as noise. Check the earnings date too, because volume spikes cluster around reports.
How often should I recheck dark pool data for a stock I own?
Once a week is enough for weekly data, each time FINRA publishes a new Tier 1 week. Recheck short interest twice a month, after each settlement date, such as Oct 15 and Oct 30, 2026. Compare each reading with the prior eight weeks, not with a single earlier week.
Can dark pool data tell me if institutions are buying?
No. Dark pool data shows how many shares traded, at what price and when, but not who bought or sold. Every trade has both a buyer and a seller. To see which funds hold a stock, use Form 13F filings, which are due up to 45 days after each quarter ends.
Sources
- FINRA, OTC Transparency data: weekly ATS and non-ATS volume (retrieved Nov 11, 2026; weeks of Nov 3, 2025 to Oct 19, 2026)
- FINRA, Daily Short Sale Volume Files (consolidated NMS files, Oct 19 to 30, 2026)
- FINRA, Information Notice: key points about short sale volume data (May 10, 2019)
- FINRA, Equity short interest data (settlement date Oct 30, 2026)
- FINRA Rule 6380A, Transaction Reporting (retrieved Nov 11, 2026)
- Cboe, "2025 U.S. Equities Year in Review" (Jan 2026)
- SEC EDGAR, 10-K and 10-Q cover pages for shares outstanding (Apple, Microsoft, Nvidia, Amazon, Tesla, Coca-Cola, JPMorgan, GameStop; latest filings as of Nov 11, 2026)
- SEC, Frequently Asked Questions About Form 13F (updated Mar 6, 2026)
- Investor.gov, "Insider Transactions and Forms 3, 4, and 5" (Jan 26, 2021)
- Intrinio price history (15-minute bars) for 29 US stocks, Oct 27, 2025 to Nov 6, 2026, via TruthSayer AI's market data feed
- TruthSayer AI, Dark Pools AI performance, trade list and alerts (Nov 11, 2026)
This article is for educational purposes only and is not investment advice. Past performance does not guarantee future results.
Nizar Assanie
Nizar is the founder of TruthSayer AI and brings over two decades of experience in quantitative finance, investment analytics, and machine learning. His expertise spans systematic investment strategies, equity valuation, institutional capital flows, and predictive financial modeling. He leads the development of AI-driven investment strategies designed to identify market inefficiencies and optimize risk-adjusted returns.
