Historical Data for Backtesting: Sources, Quality and Traps

Historical Data for Backtesting

Backtests are only as good as their data. Learn data types and sources, quality checks, corporate action adjustments and how to avoid survivorship traps.

Intermediate3 min readUpdated 3 Oct 2026

Lesson 5 of 38

Every backtest depends on historical data, and flawed data produces flawed conclusions with complete confidence. Missing delisted stocks, unadjusted splits, wrong timestamps or a single bad price can turn a losing strategy into an apparent winner. Choosing the right data, understanding its limits and checking its quality are some of the least glamorous but most important steps in trading research.

Types of historical data#

Type	Granularity	Typical use	Lesson
Daily bars (OHLCV)	One bar per day	Swing and position strategies	Tick Data and OHLCV Data
Intraday bars	1 minute to 1 hour	Day trading strategies	Tick Data and OHLCV Data
Tick and trade data	Every trade	Execution research, scalping	Tick Data and OHLCV Data
Order book data	Every quote change	Market making, microstructure	Order Book Feeds: Snapshots and Incremental Updates
Fundamental data	Quarterly and annual	Value and quality strategies	Fundamental Data
Alternative data	Varies	Specialised signals	Alternative Data Explained
Corporate actions	Events	Adjusting prices correctly	Corporate Actions, Delistings and Rolls in Backtests

Data sources#

Source type	Examples	Notes
Free	Exchange websites, public APIs, some broker feeds	Often limited history, survivorship bias, errors
Broker and platform data	Data from trading platforms	Convenient; check adjustments and gaps
Commercial vendors	Providers of research grade equity, futures and options data	Costly; point in time and delisted data available
Exchange historical products	Official tick and order book archives	Highest detail; expensive and large
Crypto exchange APIs	Trade and order book history	Free but varies by exchange; venue differences

Data quality checks#

Check	What to look for
Missing periods	Gaps in dates or times
Outliers	Impossible jumps, prices outside the day's range. See Outliers and Robust Statistics
Zero or negative values	Bad records (except where negative prices really occurred)
Duplicate records	Repeated bars or trades
Timestamp consistency	Time zones, daylight saving, bar labelling (start vs end). See Timestamps, Time Zones and Daylight Saving
Corporate actions	Splits and dividends handled correctly. See Splits and Dividends in Price Data
Cross checks	Compare against a second source

Point in time data#

Fundamental data and index memberships must reflect what was known on each date, not later revisions or today's lists. Point in time databases preserve the original values. See Point-in-Time and Survivorship-Free Data and Look-Ahead Bias.

Futures and options data#

Futures: each contract expires, so long histories require continuous series built with roll rules and adjustments. See Continuous Futures and Back-Adjustment.
Options: many strikes and expiries; bid and ask quotes matter more than last trades, which can be stale.

How much history?#

Enough to include different market regimes: bull and bear markets, high and low volatility, rising and falling rates. For daily strategies, 10 to 20 years or more is common; for intraday strategies, several years of high quality intraday data may suffice, but should still cover varied conditions. See Structural Breaks and Regime Changes.

Data storage and versioning#

Store raw data unchanged, apply cleaning in documented steps and version datasets so backtests can be reproduced. See Data Storage, Compression and Caching and Data Versioning, Lineage and Schemas.

Frequently asked questions#

What data do I need for backtesting?#

Price data at the right frequency, adjusted for corporate actions, plus any fundamental or alternative data your strategy uses, all free of survivorship and look ahead bias.

Is free data good enough for backtesting?#

Sometimes for simple daily strategies, but free data often has errors, gaps and survivorship bias, so it should be checked carefully.

What is survivorship bias in data?#

Using only securities that exist today, which excludes failures and makes historical performance look better than it was.

Next, learn how to test on data the strategy has never seen in In-Sample vs Out-of-Sample Testing.

Check your understanding

3 quick questions on this lesson. Get them all right to finish it.

Turn on JavaScript to take the quiz.

Finished this lesson?Sign in to save your progress across devices.

Next lessonIn-Sample vs Out-of-Sample TestingOut of sample testing checks a strategy on data not used to build it. Learn train, validation and holdout splits, common mistakes and how to read results.