Imagine testing a stock strategy from 2000 to today using only companies still listed today. You’ve given the strategy information it couldn’t have known: which companies would survive. That’s survivorship bias, and it can make a weak idea look much stronger than it was.
What survivorship bias is
Survivorship bias happens when a backtest includes only the companies or funds that survived to the present. Companies that went bankrupt, were acquired or were delisted for poor performance quietly disappear from the data. The test omits opportunities and losses from companies that disappeared, including potential holdings the strategy would have selected.
Why it inflates backtest results
- Losers are missing. Missing delisted stocks remove losses the strategy might have taken.
- Hindsight sneaks in. Today’s index members were partly chosen because they did well, which is information you would not have had at the time.
- Risk looks lower. Without the blow-ups, drawdowns appear shallower than they really were.
A simple example
| Survivor-only test | Test with delisted stocks | |
|---|---|---|
| Universe | Companies listed today | Eligible companies on each historical date, including later delistings |
| Bankruptcies | Not included | Included, with their losses |
| Typical result | Higher returns, smaller drawdowns | More realistic returns and drawdowns |
How to avoid survivorship bias
- Use data that includes delisted symbols. This is the single most important fix.
- Use point-in-time universes. If a strategy trades index members, use the index as it was on each date, not as it is today.
- Be skeptical of results that look too smooth. Exceptional backtests on survivor-only data deserve a second look.
- Test across many symbols. A strategy that only works on a handful of famous winners is a warning sign.
Where survivorship bias hides
The obvious case is a stock list built from today’s index members. But the bias can creep in through less obvious doors too:
- Free or cheap data sources. Many only carry symbols that still trade, because delisted histories are harder to collect and maintain.
- Watchlists built from memory. Testing a strategy on the famous winners you already know is survivorship bias with extra hindsight.
- Fund and ETF universes. ETFs and mutual funds close or merge too. A test that only includes funds still open today ignores the ones that were shut down after poor results.
- Reporting only research winners. Testing fifty ideas and reporting only the best three creates a related selection bias. Keep the failed trials too, so you can see how much searching produced the result.
How much difference can it make?
The size of the effect depends on the strategy. A strategy that holds large, stable companies for years may be only modestly affected. A strategy that buys small, volatile or beaten-down stocks can be affected much more, because those are exactly the companies most likely to be delisted. Strategies that buy stocks after big drops are especially exposed: a survivor-only test omits companies that fell and then disappeared. A stock can survive without recovering, but removing failures still distorts the pool.
That is why the same rules can look excellent on one data set and ordinary on another. Compare the two tests with the same rules. If the apparent advantage depends on leaving out failed companies, you have found a data-selection problem to resolve before trusting it.
Survivorship bias and other backtest traps
Survivorship bias is one of several traps that make backtests look better than live trading. The others usually travel with it:
- Look-ahead bias. Using information in a test that would not have been available on the trading date, such as revised financial data.
- Overfitting. Tuning rules so closely to past data that they capture noise instead of a repeatable pattern. Walk-forward testing is one of the best defenses.
- Ignoring costs. Commissions, slippage and the bid-ask spread add up, especially for strategies that trade often.
A quick checklist before you trust a backtest
- Does the data include companies that were delisted, acquired or went bankrupt?
- Was the trading universe defined as it existed on each date, not as it exists today?
- Does the test cover more than one market cycle, including at least one deep bear market?
- Are trading costs included?
- Does the strategy still look reasonable when its settings change slightly?
Use any missing answer as a concrete item to investigate before deciding what the test supports.
Why delisted data is hard to get
Keeping a clean record of every company that ever traded is expensive. Symbols get reused, companies merge and change names, and price histories for failed companies are often incomplete or adjusted incorrectly. Many data vendors simply drop delisted symbols to keep their databases tidy. That is why it is worth asking any data provider directly whether delisted companies are included and how far back the coverage goes. Ask for the coverage details rather than guessing from a vendor’s headline symbol count.
Survivorship bias outside the stock market
The same effect shows up in fund rankings, trading newsletters and even trading strategies shared online. You mostly hear about the ones that worked. The ones that failed quietly disappear, which makes the survivors look more impressive than they really are. Keeping a record of every idea you test, including the failures, is a simple way to keep your own research honest.
Check the actual universe in Portfolio Boss
Portfolio Boss includes delisted data. Its Instruments Panel shows listed and delisted counts for the portfolios you add, so you can inspect the universe you’re testing. For an index strategy, historical membership is a separate requirement: having delisted prices in the database doesn’t make today’s member list a historical one.
PB’s Backtest Panel lets you set research dates and separate in-sample and out-of-sample periods. Use those controls to keep the universe, available history, and evaluation period clear.
Founder Dan Murphy began trading in 1997 and built Portfolio Boss to turn market research into a practical trading process. Let the software handle repetitive strategy-building work while you review the evidence that makes a candidate worth keeping.
Put your next idea to the test, free
Build your first strategy with market data included, then inspect the universe and test dates behind the results. Portfolio Boss Evo is 100% free for life for the first 100,000 signups ($1,999 value). Download now. Advanced modules are paid.
Backtested and hypothetical results have limitations and do not guarantee future results.
