Research
Research begins with a question, not a backtest result.
Most of what looks like signal is noise fitted after the fact. Mizanook’s research process is built to make that mistake hard to commit — and easy to detect.
A failed strategy is a result.
Method
Eight commitments that hold across every study, in order.
- 01
Hypothesis
Research begins with an economic question — a reason a pattern might exist — not with a backtest that happened to look good.
- 02
Pre-registration
The rules, universe, costs and evaluation windows are fixed in advance and recorded, so results can't be reverse-engineered from the answer.
- 03
Point-in-time data
Decisions use only information that was actually available at the time. No future prices, no revised fundamentals leaking backward.
- 04
Transaction costs
Fees, spread, slippage and turnover are modeled conservatively. A strategy that only works at zero cost does not work.
- 05
Out-of-sample testing
Ideas are judged on data they were not developed on. Fitting the past is easy; that is precisely why it doesn't count.
- 06
Multiple-testing controls
When many ideas are tried, the bar rises accordingly. We account for the number of things tested rather than celebrating the luckiest one.
- 07
Prospective testing
Where a clean historical holdout no longer exists, evidence must be earned forward, in real time, before anything is trusted.
- 08
Production separation
Research and live systems are separated. A promising study does not silently become a deployed strategy.
What we don’t publish
This page describes method, not positions. Mizanook does not publish proprietary strategy parameters, private signal definitions, credentials or account information. Describing how research is disciplined is useful; disclosing the specifics is neither necessary nor prudent.
Nothing on this page is a performance claim. Research results, where they exist, are treated as evidence about method — not as promises of future return.