Finding a repetition in your history does not mean you have discovered a law of your trading. It may be a useful pattern, but it may also come from a small sample, retrospective classification or a combination found after trying many possibilities. Serious analysis begins when an impression becomes a defined, comparable hypothesis that is allowed to fail.
A pattern deserves attention when it defines the opportunity, preserves the denominator, survives a reasonable comparison and can be checked with new data.
Start with a hypothesis that can fail
Avoid verdicts such as “I trade worse in the afternoon”. Specify behaviour, condition and measure: “Between 16:00 and 18:00 my entries are farther from the planned level than in comparable sessions”. A precise statement can be supported, weakened or rejected.
Preserve the original trading plan rule and define the finding before trying new splits.
If you study behaviour after a loss, specify how much later activity belongs to that condition. The more flexible the definition, the easier it becomes to collect favourable examples.
Count opportunities, not only trades
The correct denominator is the number of occasions on which the behaviour could occur. If confirmation was required for twelve entries and you skipped it seven times, the frequency is seven out of twelve. The other two hundred trades in your history do not automatically belong to that question.
Define a comparable opportunity through asset, setup, session, direction, volatility or any necessary condition. Do not add filters simply because they improve the story after seeing the result.
Apply four filters before calling it evidence
Review sample, frequency, context and impact. Sample is the number of comparable opportunities; frequency is the proportion showing the behaviour; context shows where it concentrates; and impact describes what actually changes when it appears.
Impact should not mean profit alone. It can include risk, entry distance, adherence, waiting time or execution quality. A frequent but immaterial repetition may describe the environment without justifying a new rule.
- Sample: opportunities under the defined condition.
- Frequency: observed cases divided by opportunities.
- Context: asset, setup, time, volatility and sequence.
- Impact: practical difference from the baseline.
Compare against a baseline
Knowing that risk per trade increased in six sessions is not enough. Compare those sessions with ones where the condition was absent and with your planned level. Behaviour may look special when it is actually common across your trading.
Example: in 8 of 12 afternoon opportunities you entered before confirmation, compared with 9 of 38 morning opportunities using the same setup. That comparison describes a difference; it does not yet explain the cause or guarantee that it will persist.
Avoid retrospective labels and moving criteria
Classifying a trade after knowing the outcome lets the criterion bend towards the conclusion. Whenever possible, track emotions, setup and invalidation close to the decision and preserve the original note.
If the criterion is new, write an observable definition before reclassifying: “entry more than 0.25R above the planned level” is more reproducible than “FOMO entry”. Document changes in definitions so periods using different labels are not treated as equivalent.
There is no magic sample size
Twenty trades do not automatically turn an impression into evidence, just as one hundred poorly defined records do not repair a flawed comparison. The required sample depends on event frequency, variability, the size of the difference and the precision you need.
Proportions based on few cases contain substantial uncertainty. Always show numerator, denominator and period instead of a scientific-looking percentage alone. Personal decisions can use provisional evidence, but it should be described honestly.
If a setup is rare, extend the period or combine only conditions that remain equivalent. Mixing different opportunities to inflate the count creates decorative precision.
The more combinations you search, the more coincidences you find
Crossing assets, times, weekdays, labels, outcomes and emotional states creates many possible hypotheses. Explore enough cuts and one will look exceptional by chance. Selecting only the most striking one after seeing it exaggerates the strength of the finding.
Separate exploration from confirmation. Use part of the history to discover the hypothesis and reserve later cases to test it with the same criteria. Record how many variants you inspected; the winner of an invisible competition should not be presented as the only question asked.
Your journal and backtest do not need to become a statistics textbook. It only needs to preserve the path to the discovery and keep the conclusion proportional to the available evidence.
Correlation does not establish the cause
Early entries occurring in the afternoon do not prove that the clock is responsible. Volatility may rise, you may have spent longer at the screen, switched assets or arrived after a sequence of results.
Break down alternative explanations: compare the same setup across times, the same session duration at other times, and afternoons with and without recent losses. If you cannot isolate the mechanism, keep the association as a monitoring signal rather than certainty.
A relationship can still be useful before its cause is known. It may justify an extra check, but not a permanent identity label or a promise about future results.
Turn the finding into a prospective test
Define a limited intervention and change one variable. If the hypothesis is that you enter too early after several hours in a session, require written confirmation for the next twenty comparable opportunities while keeping other rules stable.
Choose the measures in advance: adherence, entry distance, risk, result in R or another metric. Include cases that contradict the hypothesis and compare with the baseline. A weekly review should assess the test, not redesign it whenever an inconvenient result appears.
If the difference does not repeat, the work was still valuable. You avoided turning an impression into a permanent restriction.
Keep evidence reviewable with TRAZZA
In the TRAZZA trading journal, evidence preserves the hypothesis, definition, sample, period, baseline, impact and tested action. That traceability shows how a conclusion was reached and which new data strengthens or weakens it.
Markets, strategies and routines change. Add a review date and state: exploratory, provisional, confirmed in a later sample or rejected. An old finding should not become eternal truth simply because it survived inside a label.
The goal is not to collect patterns. It is to focus on frequent, relevant and changeable behaviour while discarding coincidences that only add noise.
Review checklist
- ✓I defined behaviour, condition, measure and period.
- ✓I counted comparable opportunities and preserved the denominator.
- ✓I compared the finding with a baseline.
- ✓I checked alternative explanations and retrospective criteria.
- ✓I separated discovery data from later confirmation.
- ✓I recorded cases that contradict the hypothesis too.
Frequently asked questions
How many trades do I need to identify a trading pattern?+
There is no universal number. It depends on event frequency, variability, the difference you want to detect and the precision required. Always show cases, opportunities and period, and treat small results as provisional evidence.
How can I tell whether a trading pattern is a coincidence?+
Define the hypothesis before searching further, use comparable opportunities, compare with a baseline and check whether the relationship repeats in later data. The more combinations you explored, the more caution the finding requires.
Should I analyse trades or every opportunity?+
The denominator should include occasions when the rule or behaviour could apply. Counting only selected trades may hide skipped signals and distort the true frequency.
Does correlation explain why a trading pattern occurs?+
No. An association may coexist with time, volatility, fatigue, setup or a result sequence. Examine alternatives and describe it as a relationship rather than a cause until the mechanism can be isolated.
How do I validate a pattern found in my trading journal?+
Fix the criteria and intervention before observing new cases, reserve a later sample, change one variable and compare against the baseline. Record confirmations and contradictions, then review whether the finding is still current.