01
Research question
Clearly defined before testing.
Quant ResearchDesk Mandatory research protocolVERSION 1.0 / ADOPTED 29 SEPTEMBER 2026
One standard.
Every serious test.
QRD Research Standard v1.0 is the default mandatory framework for future quantitative research, backtests, validation studies and published research papers.
Core principleA backtest is evidence only when the research process makes it difficult for bias, unrealistic execution, data leakage and parameter overfitting to create a false result.
01Specify before testingQuestion, rationale, universe, data
02Model executionCosts, slippage, borrow, impact
03Validate chronologicallyWalk-forward, pure OOS, holdout
04Freeze before forward OOSNo retrospective rule changes
Mandatory research checklist
The controls behind
every QRD result.
Use the categories below to inspect the mandatory controls. The purpose is comparability: every strategy family should face the same research discipline rather than a custom standard invented after results are known.
Mandatory controls
Design & data
Define the question, economic rationale, universe and information set before interpretation.
02
Economic hypothesis
Explain why the inefficiency might exist and why it may persist.
03
Asset universe
Document the exact selection procedure and eligibility rules.
04
Data
Disclose vendor, frequency, date range, adjustments and relevant data handling.
05
Point-in-time integrity
Document when each input became available to the strategy.
06
Survivorship bias
Test for it where possible and disclose any remaining limitation.
07
Signal timestamp
State the exact moment the signal becomes known.
Evidence classification
Status is evidence quality,
not marketing language.
Every completed research programme ends with an explicit status. A study should not be labelled VALIDATED unless the required controls are satisfied and documented.
01
FAIL
The hypothesis does not survive realistic costs, OOS testing, robustness checks or predefined failure criteria.
02
INCONCLUSIVE
Evidence is mixed, unstable, too weak, too data-limited or too regime-dependent to support advancement.
03
CANDIDATE
The result passes the principal research gates and is sufficiently robust to justify final validation and/or forward testing.
04
VALIDATED
The frozen strategy has passed the required OOS/holdout, robustness, cost and statistical controls defined for the study. Validation does not imply future profitability.
Research maturity
Do not confuse discovery
with forward evidence.
Once a finalist is selected for forward testing, signal logic, parameters, universe rules, execution assumptions and portfolio/risk rules are frozen and assigned a reproducible version.
01Discovery
02Candidate
03Validated
04Frozen
05Forward OOS
06Live observation
07Retired
Freeze means freeze.
Material rule changes create a new strategy version and restart the forward-observation clock.
Forward results stay separate.
Forward performance must be tracked separately from the historical backtest.
Validation is not a promise.
Validation describes the research process. It does not imply future profitability.
Language discipline
Explore methodology Control, measure and reduce overfitting risk.
QRD should not claim that a backtest has “no overfitting”. Historical research cannot guarantee that overfitting is absent; the standard is designed to make the risk visible and harder to hide.