Datasets:
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 Daily Price
This dataset includes daily price data for various stocks.
41,554,394 rows over 20,723 symbols, 9 columns, covering 1962-01-02 to 2026-11-04. Refreshed monthly.
Strategies Built on This Data
2,401 papers in the Papers With Backtest catalogue declare this dataset as an input. 2,238 of them have been coded and run over their own full history. The median replicated Sharpe ratio is +0.35, and 45% clear a t-statistic of 1.96 on their own sample, against 48% across all 4,837 replications in the library.
Some of the strongest results that lean on it:
| Strategy | Sharpe | t-stat | Tested over |
|---|---|---|---|
| Important Characteristics, Weaknesses and Errors in German Equity Data from Thomson | +1.68 | 10.0 | 36 years |
| Understanding Momentum and Reversal? | +1.36 | 8.1 | 35 years |
| Analytical Solution for Kelly’s Criterion for Multiple Outcomes | +1.31 | 7.8 | 35 years |
| End-to-end Large Portfolio Optimization for Variance Minimization with Neural Networks Through Covariance Cleaning | +1.28 | 7.2 | 32 years |
A Sharpe ratio quoted without its t-statistic hides how much of the library cannot be distinguished from zero, which is why both are shown. The figures are in-sample over each strategy's own window and carry no transaction costs.
Why It Matters
This dataset supports equity signal research and portfolio construction by:
- Signal research: Long history of OHLCV data powers momentum, mean-reversion, and breakout studies across equities.
- Corporate actions adjusted: Adjusted closes support total-return analyses and bias-free backtests.
- Liquidity screens: Volume trends help design tradability filters and slippage assumptions.
Reused tickers. A ticker that designated several instruments over the years does not stitch them into one series: the instrument trading under it most recently keeps the bare ticker, and each earlier one keeps its own rows under <TICKER>_<YYYYMMDD>, the last day it traded under that ticker. ASO is Academy Sports (2020-), ASO_20061103 is AmSouth Bancorporation (2003-2006). Load either name like any other symbol.
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-Daily-Price", symbols=["AAPL"])
print(df.iloc[0, :])
Example Output:
symbol AAPL
date 1980-12-12
volume 469033600
close 0.098389
open 0.098389
high 0.098817
low 0.098389
Columns
| Column Name | Description |
|---|---|
| symbol | Stock ticker. |
| date | Trading date (YYYY-MM-DD). |
| volume | Trading volume. |
| close | Closing price for the session. |
| open | Opening price for the session. |
| high | Highest traded price during the day. |
| low | Lowest traded price during the day. |
| adj_close | Closing price adjusted for splits and dividends. load_dataset rescales the other three prices by adj_close / close and returns it as close, so the snippet above prints adjusted prices and no separate column. Pass adjust=False for the raw ones. |
| type | Type code of the instrument the row belongs to, read from a point-in-time instrument record: CS common stock, ADRC American depositary receipt, OS ordinary share, GDR global depositary receipt, NYRS New York registry share. Null when no record ever typed the instrument (see below). get_pricing does not return it; load_dataset does. |
Instrument Type
Every row says which kind of instrument it is (type). The code is the one stated for the instrument over the period it traded under that ticker, so a reused ticker carries the type of each instrument on its own rows, and an adjusted series carries the type of the instrument it prices today.
Measured on the build of 16 November 2026: 88.3% of the rows carry a type (CS 83.0%, ADRC 5.0%) and 11.7% are null. The nulls sit where the record is thin: 1.4% of the rows before 2003 (a series priced under its current ticker carries its current type over its whole history), 34% in 2003-2007, 17% in 2008-2015, 1.4% since 2016. A few rows carry a code of another kind and stay in this panel: operating companies the record types FUND, mostly real-estate trusts such as DHC (84,091 rows), and listed partnerships typed UNIT (32,028); the code says what the record says, the repository what the instrument trades as.
What a null means. The instrument record only types instruments from the mid-2000s on, and almost never before 2008. A null row is one whose class was decided by default: nothing classed it as a fund, preferred share, warrant or structured product (by its record, its name, or its ticker's notation, such as a lower-case p for a preferred series), so it is filed as a stock. Most are operating companies that were delisted before the record existed; filter on type.notna() for a panel restricted to positively identified stocks, at the cost of most of the pre-2008 history.
Stray prints are not published. A run of fewer than five bars that no instrument record places, more than a year away from the real series under the same ticker, is not a price history: 81 such series across the five equity repositories, 96 rows in all (52 of them single bars printed on 2007-06-13 under tickers long dead), are left out rather than published under a name of their own.
What This Data Does Not Cover
Survivorship. Cross-checked against AlphaVantage-Stocks-Daily-ListingStatus, which lists 16,675 US symbols of which 7,276 are delisted, only 3.5% of the symbols in this file are companies that have since been delisted. On the listing record, 44% of the US common stocks that ever traded are. The gap comes from the vendor feeds, which serve currently listed symbols, and it is not a pipeline error, but the consequence for a backtest is the same: a cross-sectional sort run on this file is a sort over the survivors, and its Sharpe ratio is flattered by that.
That universe now exists: Stocks-Daily-LiquidUniverse states, for each session since 2003, which large liquid names were actually trading, and 35% of the 2,013 names that pass through its top 500 are delisted today. pwb-toolbox resolves symbols=["sp500"] to it and keeps each name to its own window, so a cross-sectional sort run through the loader is no longer a sort over the survivors. Loading this file by an explicit symbol list still is.
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
- Dataset page and coverage charts
- The strategy catalogue, 3,806 papers and 4,837 replicated strategies
pwb-toolbox, the loader used in the snippet aboveawesome-systematic-trading, the replicated strategies with their measured Sharpe- Every dataset in this organisation
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
- 7,884