Key idea
A pattern discovered after seeing the outcome needs a separate test before you treat it as a reliable research rule.
After a dramatic price move, it is easy to find a chart feature or data point that seemed to announce it. The harder question is whether the same feature would have attracted your attention before the move—and how often it appeared without that outcome.
A useful pattern study separates discovery from evaluation. Discovery generates a hypothesis. Evaluation applies a fixed definition to examples that did not shape the hypothesis.
Write the rule before collecting the evaluation sample
Specify the universe, observation time, required fields, and qualifying conditions. Avoid a rule that depends on adjectives such as clean, strong, or obvious unless you define how those judgments are made.
For a hypothetical study, you might investigate whether a particular relative-performance condition persists to a later checkpoint. Write both the condition and the checkpoint before reviewing the evaluation examples. Do not choose the day's best exit afterward.
Preserve the full selection trail
Record every qualifying example, including those with no move, missing data, or an unfavorable result. Exclusions should follow a stated rule and remain visible. Removing inconvenient cases after seeing them can make a weak pattern appear stronger.
Also retain the date the rule was created. If you revise it after examining results, the revised rule begins a new exploratory cycle. The original evaluation should not quietly inherit the improvement.
Use information available at the observation time
Historical datasets can contain later revisions, finalized values, or classifications that were not available in real time. Check whether the inputs match what a researcher could have known at the chosen cutoff.
If that cannot be established, label the study accordingly. You can still learn from a historical relationship, but you should not describe it as a demonstrated real-time decision process.
Separate market behavior from execution results
A later price move is not the same as an achievable return. Options studies also need appropriate contract quotes, spreads, fees, and fill assumptions. Without those, keep the conclusion about the observed pattern rather than hypothetical profitability.
Try a small exercise with ten unseen examples after defining the rule. The goal is not statistical proof from a tiny sample. It is to learn whether your definition can be applied consistently and whether the data is sufficient.
DAXA's research tools can help you collect observations, but the credibility of a study depends on the rules you preserve and the cases you allow to challenge your hypothesis.
Bring more context to your research
Explore DAXA’s research workspace and choose a membership that fits the information you use.