1
Survivorship bias
Exclude failed instruments and historical performance usually looks stronger.
Lesson · Backtesting and Trading Systems
Survivorship, Selection, and Corporate Actions
Build a point-in-time universe, include failed and delisted instruments, and adjust splits and distributions without using information learned later.
Core concepts
What to understand
Keep each definition operational so it can be implemented and tested.
2
Selection bias
Choosing markets or periods after reviewing results changes the question being tested.
3
Corporate actions
Splits, dividends, mergers, and delistings must be handled consistently with the strategy.
Interactive lab
Universe and Corporate Actions Lab
Toggle point-in-time membership, delisted symbols, and split adjustment. Compare the honest portfolio with the biased version.
Universe and Corporate Actions Lab
Deterministic synthetic educational data · no live market connection
BACKTEST MODEL
Process
Reconstruct the investable set
Universe membership is data, not a static list.
Date membership
Record when each instrument entered and left the universe.
Include failures
Retain delistings, bankruptcies, and unavailable periods.
Adjust correctly
Separate price adjustment from actual cash distributions and trade execution.
Knowledge check
Check your understanding
Select one answer for each question. Feedback appears immediately.
1. Testing only companies that exist today creates:
The test removes historical failures that were investable at the time.
2. A 2-for-1 split should generally cause an unadjusted price series to:
Without adjustment, the mechanical price change looks like a large return.
3. Point-in-time universe data prevents the test from:
The system should only know constituents available on that historical date.
4. SPLT is quoted at 111.33 and splits two-for-one to 55.66. Adjusted, the step return for that period is:
55.66 × 2 ÷ 111.33 − 1 = 0%. Unadjusted, the lab books −50% for that name, which is −50% ÷ 8 = −6.25 points on the equal-weight portfolio's step.
Practical rules
- Use point-in-time constituent membership.
- Include delisting returns or a documented conservative treatment.
- Verify whether vendor prices are raw or adjusted.
Common mistakes
- Downloading today’s index members for a ten-year test.
- Removing bankrupt symbols as bad data.
- Mixing adjusted prices with unadjusted cash flows.
References
Official reading
The labs are simplified simulations. These sources cover historical performance, data, costs, or execution limitations.
FINRA · Smart Beta—What You Need to Know
Explains that historical backtesting is hypothetical and does not predict future performance.
CME Group · Historical Market Data
Describes historical market data used for market analysis and strategy validation.