01
Hypothesis
State the research question and economic rationale before interpreting results.
Quant ResearchDesk Research methodologyQRD / EVIDENCE SYSTEM / 01-07
Evidence before assertion.
QRD separates exploration from validation, historical evidence from prospective evidence, and an attractive backtest from a strategy that has survived a documented research process.
Read the mandatory standard02
Experiment
Define the universe, data, signal timing, execution assumptions and complete search space.
03
Walk-forward OOS
Respect chronology. Select and test without leaking future information into model choice.
04
Robustness
Stress costs, parameter neighbourhoods, regimes, data choices and sampling uncertainty.
05
Statistical validation
Use holdouts, Deflated Sharpe Ratio, PBO/CPCV and bootstrap or Monte Carlo where appropriate.
06
Freeze
Lock signal logic, parameters, universe, execution and portfolio rules into a reproducible version.
07
Forward OOS
Track the frozen model prospectively and keep forward evidence separate from historical backtests.
Research discipline
Make the evidence trail
hard to rewrite.
The strongest research system is not the one that produces the smoothest equity curve. It is the one that records what was asked, what was tested, what failed and what remained unchanged after freeze.
Point-in-time data
Record when information became available rather than when it was later stored.
Realistic execution
Costs, spread, slippage, market impact and borrow assumptions belong inside the research design.
Search transparency
Report the entire search space and experiment count, not only the winning configuration.
Visible failure
A failed hypothesis is a legitimate research outcome and should remain part of the evidence trail.
Publication minimum
A technically competent reader should know what happened.
Every QRD paper should disclose enough information to understand what was tested, when decisions were made, how trades were simulated, how many alternatives were searched, which robustness gates were applied, what failed and what limitations remain.
01Research question02Economic rationale03Data and universe04Execution assumptions05Transaction costs06Search space07Pure OOS results08Robustness tests09Failed configurations10Limitations11Version / hash12Forward status
Research governance
Explore research When rules change, the version changes.
A frozen strategy should not be retrospectively rewritten. Material changes create a new version so the historical record remains interpretable under the standard that governed it.