Look-ahead bias: when was the signal actually knowable?

Marker location and knowledge time

Suppose a swing high requires two later bars with lower highs. The candidate high is visible immediately, but its confirmed status is not. A historical chart may place the completed marker over that earlier high for readability. Trading as though confirmation existed there gives the test advance knowledge.

The same issue occurs when a rule uses the final daily close but assumes execution at that exact close without a feasible submission process. Financial reports must be dated by public availability rather than fiscal period end, and later revisions must not silently replace the information originally available.

Worked example: a Wednesday high confirmed Friday

Use five fictional daily highs: Monday 100, Tuesday 103, Wednesday 108, Thursday 106 and Friday 105. Wednesday becomes a confirmed two-bars-later swing high only after Friday's bar closes. A strategy that needs this confirmation cannot legitimately enter on Wednesday using that label.

Assume Wednesday closed at $107, Friday at $104 and the next session opened at $102. A short test entered at $107 and covered at $100 would record $7 per share. An assumed next-session entry at $102 would record $2 before costs, if borrowing and execution were actually available. The $5 difference is a timing assumption, not improved forecasting.

A bar does not contain the full event sequence

Daily bars do not reveal the order of every intraday event. If a bar touches both a target and a stop, its high and low alone cannot establish which was reached first. Time zones, release delays and data revisions also matter. A clean-looking chart is insufficient evidence of feasible execution.

Write three timestamps for each signal

On available closed daily, weekly or monthly bars, list the observation time, confirmation time and earliest assumed action time for each manually studied signal. Check every condition using only bars available by that decision. Student Wyckoff Pro can support closed-bar analysis; this timestamp audit does not imply an automated historical execution simulation.

Primary sources

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

• Investment Model Validation: A Guide for Practitioners

Sources checked: 2026-10-23