Robustness testing · TradersWeek

Skip to main content

Output: Signal
Input: Rule inputs, market state and execution constraints
Horizon: The system holding period
Best use: Make entry, exit and abstention rules testable as one procedure.

Input: Rule inputs, market state and execution constraints.
Output: signal.
Use boundary: Make entry, exit and abstention rules testable as one procedure.

The graph3

How the method is built, and where it has been read

Depends on

Often reviewed together

Mechanical trading system18 readings

Often reviewed together

Walk-forward analysis24 readings

Often reviewed together

System optimization26 readings

Read in the archive24 of 80 readings

1986Degrees of freedom in trading system optimizationSystem design 1988Walk-forward and neighborhood tests after optimizationSystem design 1988Undisclosed rules block system robustness testsSystem design 1988Isolate nested formulas before judging signalsSystem design 1988Testing re-optimization calendars against random parameter controlsSystem design 1989Binary search limits on multi-peak average gridsSystem design 1989Known-bias price streams for system evaluationSystem design 1989Parameter neighborhoods that survive a shiftSystem design 1990Use profit mapping to keep a cycle and stop plateauSystem design 1990Why popular indicator optimization fails robustnessSystem design 1991Retesting weighted indicator balances across horizonsStatistical tests 1991Stop bounds versus added system parametersStops and loss limits 1992Multi-year evaluation of MAE-bounded mechanical rulesSystem design 1992Constructing forecast models with regression, walk-forward, and robustnessTime series and forecasting 1992Diagnose regimes before you lock parametersSystem design 1992When stops change system timingSystem design 1993Walk-forward halt rules for forecast modelsTime series and forecasting 1994Walk-forward evaluation of genetic index rulesSystem design 1995Critiquing neural nets as incomplete trading systemsSystem design 1995Input pruning as walk-forward system evaluationTime series and forecasting 1996Rebuild the equity-path ratio before it ranks a designed systemSystem design 1996Parameter grids can fit random walksSystem design 1996Walk-forward analysis belongs in the design of a mechanical trading systemTime series and forecasting 1997When a holdout fails, discard the rule setSystem design

This registry entry describes a research method. It is not investment advice and does not make a performance claim.