Research as a discipline

Good backtesting lives inside a good research process; without one, a researcher tortures the data until something fits and calls it an edge. These pages lay out the workflow professionals use — from generating and framing an idea as a testable hypothesis, to defining rules and choosing parameters honestly, validating, reviewing performance and improving without overfitting. The process is what keeps discovery separate from self-deception.

Research Process: The quantitative research process is the disciplined workflow that turns a trading idea into a validated — or rejected — strategy while guarding against overfitting and bias. Its stages are idea generation, framing the idea as a falsifiable hypothesis, defining explicit rules, selecting parameters with restraint, running a validation workflow (out-of-sample, walk-forward, robustness), reviewing performance honestly, and improving continuously. The point of the process is to keep genuine discovery separate from fitting noise.

Frequently asked questions

What is the quantitative research process?
It is the structured workflow for developing and validating a systematic strategy: generate an idea, frame it as a falsifiable hypothesis, define rules, select parameters carefully, validate out-of-sample and via robustness tests, review performance, and iterate. The discipline exists to separate a real edge from noise you have fitted.
Why start research with a hypothesis?
A hypothesis states in advance why an edge should exist — a real market behaviour like trend persistence or mean reversion. Testing a stated hypothesis is honest science; searching data for anything that worked and inventing a story afterwards is data snooping, which produces edges that vanish live.
How do I improve a strategy without overfitting?
Change one thing at a time for a stated reason, re-validate on data untouched during the change, and prefer fewer parameters and structural improvements over fitting the equity curve. Continuous improvement guided by out-of-sample and walk-forward evidence avoids the trap of tuning to historical noise.
Educational content only — not investment advice. See our Risk Disclosure.