TradeHouse Fidelity Lab · Guides

Survivorship bias in backtests

Survivorship bias appears when a backtest only includes things that still exist at the end of the test period. Instruments that were delisted, merged, went bankrupt or were voided are missing, and they are usually missing because they did badly. The backtest then looks better than any real trader could have done.

Where it shows up

• Equity universes. Using today's listed stocks to test a strategy over the past decade leaves out every company that failed in between.
• Index constituents. Testing on the current members of an index, instead of the members at each historical date.
• Funds and managers. Databases that drop closed funds overstate average returns.
• Prediction markets. Analysing only markets that resolved normally, leaving out ones that were cancelled or voided.
• Strategies themselves. Reporting only the strategies that worked and forgetting the ones abandoned. This is closer to data snooping.

How to avoid it

1. Use point-in-time universe data that includes delisted instruments.
2. Rebuild index membership as of each rebalance date.
3. Keep delisting returns, not just the last traded price.
4. For your own research, keep a log of every strategy tried, not just the survivors.

A quick test for your own data

Count the instruments in your universe at the start of the test and at the end. If the two numbers are identical over a period of several years, the data almost certainly lists only survivors, because real universes lose and gain members every year. Then check whether any instrument in your file has a final price of zero, a delisting date or a void status. A long equity backtest with none of these is a strong sign that the failures were filtered out before you started.

What a validator can and cannot see

A validator that grades your trade file can only see the trades in it. If instruments were removed before the file was exported, no tool can recover them. Fidelity Lab checks a related, visible problem: its hindsight-selection detector flags cohorts chosen because they were already profitable, where in-sample rank fails to predict out-of-sample rank. Universe construction remains your responsibility, and the report says so.

Try it on your own trades. The free Fidelity Lab tier runs one report a month on up to 5,000 legs: all 11 contamination detectors, cost re-netting and the walk-forward verdict. Run a free report or see pricing.

Frequently asked questions

How large is survivorship bias?

It depends on the universe and period. It is largest for long tests on small or volatile instruments, where failure rates are highest.

Does using an ETF avoid survivorship bias?

Testing the ETF itself avoids it for that instrument, but a strategy that selects ETFs from a current list can still suffer from it.

Related guides

• Backtest overfitting: a practical guide
• Walk-forward validation vs cross-validation
• Backtest validation tool: what it should check
• Probability of backtest overfitting (PBO)
• Trading strategy audit service
• Overfitting in backtrader
• Trading strategy robustness tests
• Monte Carlo backtests
• Look-ahead bias in backtests
• Data snooping in trading
• Out-of-sample testing in trading
• What is backtest overfitting?
• Deflated Sharpe ratio calculator
• Fidelity Lab

All buyer guides and tools

Educational content, not investment advice. Updated 2026-10-28.