Glossary
Pre-registration
Writing down a test's hypothesis, data, costs, test windows and pass threshold before any result is computed, so a failure counts as a failure.
Pre-registration means fixing the plan of a test in writing, with a date, before running it. The plan states the hypothesis, the instrument and data, the exact rules, the cost model, the development and test windows, and the threshold a result must clear to pass. After the run, the result is judged against the plan, not against what would have looked good.
Example
A plan for a gold test reads: XAUUSD, 5-minute bars, 2012–2024; rules developed on 2012–2021 and checked once on 2022–2024; costs 1.1 bp per trade; pass only if the 2022–2024 average is above 0 bp after costs, with t ≥ 2.0 over at least 200 trades. The test returns +0.2 bp with t = 0.9. It fails, and it is published as a failure.
Why it matters
Without a written plan it is easy to move the threshold, drop a bad year or try one more variant until something passes. Pre-registration makes a failed test count as a failure, which is what makes the tests that pass worth trusting.