Build a testable trading hypothesis
Goal
Turn a market observation into a claim that could be disproved.
Lesson
“This indicator works” is not a useful hypothesis. Describe the behavior, the market conditions where it should occur, and the observable signal that represents it. Name at least one condition that would make the claim fail. Avoid changing the claim after seeing results.
Separate an idea from an implementation. First write the expected behavior in plain language. Then encode it, run a small check on signal timing, and confirm that the code uses only information available at that candle close.
Practice
Write a hypothesis with a target market, timeframe, entry condition, exit condition, and a falsifying observation. Record it before the first parameter search.
Check your understanding
If a candidate loses on its frozen validation period, can you keep changing settings on that period and call it validation? No. Once you use the result to modify the strategy, that period has become development data.