Keep the claim specific
A conclusion belongs to the tested market, period, horizon and transaction assumptions. A different scope needs new evidence.
Research methods
A study starts with a question and a comparator. It ends with a recorded decision: continue, revise the hypothesis or stop.
Three principles
Design for risk
Assess drawdown, concentration, liquidity and operating failure alongside return. Preserve adverse observations.
Test implementation
Costs, turnover, borrowing, size and delay can overturn an attractive statistical result.
Testing
Define the question
State the portfolio job, comparator and result that would disprove the idea.
Fix the data boundary
Check publication times, membership, missing periods and corporate actions before constructing decisions.
Evaluate later periods
Choose settings on earlier data, then evaluate the next unseen period. Retain fold and regime variation, unsuccessful tests and trial counts.
Stress the transaction
Apply fees, spread, slippage, impact and relevant financing or borrow. Compare delayed execution and harder cost assumptions.
Observe prospectively
Freeze the decision process for paper observation; keep losses and gaps. A live pilot would add actual fills and operating consequences.
Market assumptions
| Market | Data clock | Execution issue |
|---|---|---|
| Equities | Company information and corporate actions as known on each date | Borrow, corporate actions and whole-share completion |
| Digital assets | Completed hourly observations across continuous trading | Venue liquidity, funding, fees and interruptions |
| Options | Underlying observations and timestamped option quotes | Bid/ask, multi-leg fills, margin, assignment and expiry |