NSI Research: Write the Test Before Seeing the Outcome

A compelling example is a starting point

A chart chosen after a striking move can suggest a question. It cannot by itself show how a rule behaves across ordinary conditions. NSI is presented as a proposed framework, with independent validation still necessary. ^[1] Its book can organise concepts, but publication does not establish predictive accuracy. To move from an interesting example to a testable claim, specify what the next observations would have to demonstrate before their outcomes are known.

Define the proposed observation

Write a short plan with an eligibility rule, a time window, an outcome and a comparison. For example, a hypothetical study could examine whether a precisely defined pre-decision checklist is associated with fewer departures from a written trading plan. That is a different outcome from profitability, stress reduction or market prediction. Do not quietly swap among them. Define what counts as a departure and how missing records will be handled. A private timestamped note is helpful documentation; it is not automatically a formal public preregistration.

Preserve the distinction between planned and exploratory work

The Center for Open Science explains preregistration as specifying and registering a study plan in advance, helping distinguish planned from unplanned analyses. ^[2] For an NSI research project, that principle suggests retaining the original protocol and marking later changes. An unexpected pattern can be worth exploring without being presented as a prediction made beforehand. Separate personal and market observations so that one cannot silently substitute for the other.

Count the ordinary cases too

Record all eligible observations, including no signal, missing data and unfavourable outcomes. A stopped experiment needs an explanation as much as a successful-looking one. Any study involving people or physiological data also needs appropriate consent, privacy and ethical review. This article proposes a planning discipline; it reports no completed experiment or validated NSI effect. The confidence trap provides the companion question: what evidence would change the conclusion rather than merely decorate it?