If one of these steps is weak, the system is still discretionary even if it looks automated on the surface.
State what condition you think exists: trend, mean reversion, breakout, volatility compression, or event response.
Convert the idea into entries, exits, invalidation logic, and position sizing that the creator can enforce without improvisation.
Run historical validation with realistic capital assumptions, symbols, and bad periods instead of trusting the first attractive equity curve.
Automation reduces repetitive decisions, but the operator still owns monitoring, escalation, and shutdown criteria.
Sizing is risk ÷ stop distance. Adjust the inputs and watch how many shares the same risk budget buys.
Do not automate vague pattern recognition you cannot define.