Survivorship bias: the missing losers in a backtest

Restore the historical universe

Build the eligible universe as it existed at each historical decision date. Today's index members should not be projected backwards into earlier years. Keep permanent identifiers as well as ticker histories: a renamed security or reused ticker is not automatically the same investment.

A price series ending is not a complete investment outcome. Delisting returns, acquisition payments, distributions and the timing of cash availability may matter. Missing final proceeds should be investigated and disclosed, not silently replaced with zero return or an assumption that the last quoted price was executable.

Worked example: +20% becomes −4%

Imagine ten equally weighted investments, each beginning at $100, with no rebalancing or fees. Eight finish at $120 and two fail with zero recovery. The original $1,000 portfolio ends at $960, a 4% loss. A database containing only the eight survivors would show $800 becoming $960, a 20% gain.

The 24-percentage-point difference comes entirely from sample selection. If each failed investment instead returns $10, the full portfolio ends at $980, a 2% loss. That second calculation shows why the actual terminal payment matters; delisting does not universally mean a total loss.

Delisting is not one standard outcome

Adding a few well-known bankruptcies does not prove that the dataset is complete. Eligibility rules, exchange coverage, missing histories and publication dates can still distort selection. Conversely, mergers can remove successful companies. The direction and size of bias depend on why observations disappeared and how the strategy would have treated them.

Investigate where a chart ends

When reviewing an available closed daily, weekly or monthly chart, note any history that stops abruptly and whether the symbol has changed. Treat that as a data question before interpreting a pattern failure. Student Wyckoff Pro supports closed-bar investigation; a complete historical universe and terminal investment proceeds must be verified from separate records.

Primary sources

Sources explain the mechanics and limitations; the editorial examples use invented numbers rather than market quotations.

• CRSP Monthly Stock Demo

Sources checked: 2026-10-23