What you will learn
- Setup tags
- Session and day
- Holding period
- Avoid cherry-picking tiny groups
Setup tags
Setup tags must be defined before analysis or labelled exploratory when added later. A tag invented after inspecting the best trades can select winners by construction. Preserve the distinction between hypothesis generation and evaluation.
Session and day
Session and weekday comparisons need sufficient observations and consistent timezone treatment. A high average from two trades is not directly comparable in certainty to a lower average from a much larger sample.
Holding period
Holding period can interact with financing, volatility and event exposure. Grouping by duration can be informative, but duration itself may be affected by the exit rule and outcome. Interpret it as a descriptive association unless the design supports more.
Avoid cherry-picking tiny groups
Repeatedly splitting a small dataset produces attractive subgroups by chance. Report group sizes, excluded records and the number of comparisons. Reserve later data to evaluate a selected improvement rather than declaring every favorable subgroup an edge.
Worked example
A learner tries twenty tag combinations and reports only the one with three winners. The apparent subgroup performance includes a selection process that must be disclosed.
Try it yourself
Design a follow-up evaluation for that tag without pretending the original three trades were independent confirmation.
Show the worked solution
Treat the original subgroup as exploratory. Freeze its definition, collect a later reserved sample with all eligible cases and compare under the same costs. Report the original search and the new sample's uncertainty.
Apply this to your course project
Produce a weekly report that reconciles net P&L and explains uncertainty in small groups.
Keep the calculation inputs, assumptions and decisions with your work. Practical exercises are self-reviewed; the scored knowledge checks assess the questions shown, not an independent certification of practical competence.