You need to agree to share your Hugging Face information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This dataset is free to browse and gated for download. Approval is tied to a Papers With Backtest subscription, which also covers the other datasets in this organisation and the strategy catalogue at https://paperswithbacktest.com. Plans and what each one includes: https://paperswithbacktest.com/pricing

Log in or Sign Up to review the conditions and access this dataset content.

Stocks Weekly ShortVolume

Weekly off-exchange short sale volume for US-listed stocks, from August 2009, summed from FINRA's daily Reg SHO short sale volume files.

5,233,115 rows over 16,839 symbols, 8 columns, covering 2009-08-07 to 2026-10-30. Refreshed monthly.

Why It Matters

Short volume is the flow side of short interest, and it arrives weekly rather than twice a month:

  • Pressure, not positioning: short_volume_ratio_exchange measures how much of the week's off-exchange trading was sold short. It moves before short interest does, because it counts trades rather than open positions.
  • One definition over the whole history: every week is the sum of the same four FINRA trade reporting facilities (the two Nasdaq TRFs, the NYSE TRF and the ADF), which is exactly FINRA's own consolidated file where FINRA publishes one (from August 2018). A level seen in 2010 compares with a level seen in 2026.
  • Crowding: A rising short share on falling volume is a different setup from a rising short share on rising volume, and both are visible here.

What changed in November 2026. The series used to come from a vendor and is now rebuilt from FINRA alone over its whole history, with one definition throughout. The former series did not keep one: its volumes were about twice FINRA's off-exchange figure until mid-2023 and about a third of it afterwards. Expect three differences. Volume levels differ (FINRA over former, median per stock: 0.63 for total volume, 0.60 for short volume, with a wide spread). The short share is lower (median 0.44 against 0.54) and follows the former one only loosely (median correlation over time per stock 0.58). Coverage is wider: 16,839 stocks against 6,402. Only stocks are kept: FINRA reports ETFs like stocks, and they are sorted out by the catalogue's symbol index; warrants, units, rights and preferreds the former series carried (250 symbols) are left out. A signal calibrated on the former series has to be measured again. The three participant columns are kept so that existing code runs, and are empty: FINRA does not split short volume by participant.

Load It

Installation/Upgrade:

pip install --upgrade pwb-toolbox

Load the Dataset:

from pwb_toolbox import datasets as pwb_ds

df = pwb_ds.load_dataset("Stocks-Weekly-ShortVolume", symbols=["AAPL"])
print(df.iloc[-1, :])

Columns

Column Name Description
symbol Stock ticker, as the catalogue's price tables spell it; share classes after a dot (BRK.B). ETFs, preferreds, warrants and units are left out.
datetime Friday of the reporting week (the week runs Saturday to Friday; dated the Friday even when it is a holiday).
short_volume Shares sold short during the week (short-exempt sales included), in trades reported to FINRA's facilities: off-exchange trading only.
total_volume All shares traded during the week in trades reported to those same facilities: the off-exchange volume, not the consolidated tape.
short_volume_ratio_exchange short_volume / total_volume: the short share of the week's off-exchange volume.
retail_short_ratio Empty since November 2026 (FINRA has no participant split).
institutional_short_ratio Empty since November 2026.
market_maker_short_ratio Empty since November 2026.

Source: FINRA Reg SHO daily short sale volume files.

Access

Browsing the card and the schema is open to anyone. Downloading the files needs an approved request, tied to a subscription: what each plan includes. The same subscription covers the other datasets in this organisation.

Elsewhere

Papers With Backtest publishes 32 datasets on the Hub and codes the papers that use them. Every strategy in the catalogue is run over its own full history before it is published, which is where the numbers above come from.

Downloads last month
247