Details
Where off-exchange trades come from
Dark pools are alternative trading systems (ATS) that match buyers and sellers without showing orders before the trade. Institutions use them to move size without signalling it to the lit market. Off-exchange volume also includes retail orders filled internally by wholesalers and block trades negotiated between brokers.
All of these trades still get reported. In the US they go to a FINRA Trade Reporting Facility (TRF), and the print appears on the consolidated tape. What the tape doesn't tell you is which dark pool did the trade or who was on each side. FINRA publishes ATS-level weekly volume, but with a delay of a few weeks.
In tick data you spot an off-exchange print by its venue. Our STK-TRADES schema marks it with is_off_exchange = true and a trf_id. The dark pool view (STK-DARKPOOL) adds notional_value, the NBBO at the time (bid_price, ask_price), and two flags that matter a lot: is_late and is_block.
Worked example (illustrative numbers)
Say a stock trades between $50.10 and $50.40 in the afternoon. At 15:42 ET a print of 400,000 shares at $50.18 shows up from a TRF, with the NBBO at $50.30 / $50.31 at the moment it hits the tape.
- The notional is 400,000 × $50.18 = $20.07 million, so it is a block by any standard.
- The price is 12 cents below the bid. That looks dramatic until you check
executed_timestamp. If the trade was agreed at 11:05, when the stock traded near $50.18, and only reported now, it is a late print. It tells you nothing about the price at 15:42. - If instead the execution time matches and the price sits at the midpoint, it is a normal dark pool cross. You still can't tell if the buyer or seller started it.
The useful signal is usually where blocks cluster across days. Repeated large prints near one price can mark a level where an institution keeps working an order.
Details
Filter off-exchange blocks in pandas
"""Example: off-exchange blocks from a trades file, late prints removed."""
import pandas as pd
t = pd.read_csv("trades_sample.csv", parse_dates=["timestamp"])
dark = t[t["is_off_exchange"]].copy()
dark["notional_value"] = dark["price"] * dark["size"]
blocks = dark[(dark["notional_value"] >= 1_000_000)]
blocks = blocks[blocks["session"] == "regular"]
share = dark["size"].sum() / t["size"].sum() * 100
print(f"off-exchange share of volume: {share:.1f}%")
levels = (blocks.assign(level=blocks["price"].round(1))
.groupby("level")["notional_value"].sum()
.sort_values(ascending=False).head(5))
print(levels)
Read the first number as how much of the day's volume printed off exchange. Treat levels as the price levels where the most block notional traded. If the same level shows up on several days in a row, mark it on your chart. With STK-DARKPOOL rows, drop is_late prints before grouping.
Watch out
Common mistakes
Treating every dark print as a buy. A print has two sides. Without the order that started it, a big trade is not bullish or bearish on its own. Comparing the price to the NBBO at execution gives a weak hint at best.
Mixing up the report time and the trade time. Late and out-of-sequence prints are common in block trading. Always compare the execution time with the report time before you read anything into the price.
Counting retail internalisation as institutional. Wholesaler fills are off-exchange too, and they are mostly small retail orders. Filter by size or notional if you are looking for institutional activity.
Reading weekly ATS data as real time. FINRA's ATS breakdown is published weeks later. It is good for research on who trades a stock, not for intraday signals.
Details
Get the data
Off-exchange prints, filtered from the full trade tape. What you'll get: every TRF print with
trf_id,notional_value, NBBO at the time, andis_lateandis_blockflags.
- Dark pool data, available in early access and planned for Pro ($79/mo)
- Also inside our stock tick data files: filter on
is_off_exchange- We don't have a free US stock trade sample yet. A sample file is available on request.
FAQ
Frequently asked questions
What is dark pool trading?
Dark pool trading is buying and selling through private venues that don't display orders before a trade executes. The trades are still reported to the public tape afterwards. Institutions use dark pools to trade size with less price impact.
Are dark pool trades reported?
Yes. In the US, off-exchange trades are reported to a FINRA Trade Reporting Facility and appear on the consolidated tape. The tape shows the price and size but not which dark pool executed the trade.
Do dark pool prints predict price moves?
Not reliably on their own. A single print has a buyer and a seller and often arrives late. Some traders track price levels where large blocks repeat over several days, but that is a context tool rather than a signal.
Why do some dark pool prints show odd prices?
Usually because they were reported late or were negotiated at a price agreed earlier. Check the execution timestamp and any late or out-of-sequence condition codes before reading the price.
Keep reading