Combining Multiple Objectives and Constraints in Trading Strategy Optimization

Summary

The article presents a generic formulation for optimizing trading systems when a platform accepts bounds on input variables but cannot directly enforce other constraints. It combines multiple performance objectives—such as return, Sharpe ratio, recovery factor, or win rate—into a weighted sum. Each objective is divided by a user-chosen target to put measures on comparable scales, and the weights express their relative importance.

Additional constraints are handled through penalty terms: violations reduce the combined objective, while feasible designs receive an adjustable offset. The article explains how these values influence a genetic optimizer and its ranking of candidate parameter sets. It distinguishes that approach from exhaustive evaluation and discusses choosing targets and bounds, relaxing conflicting constraints, and reviewing candidates by other metrics. The method is a practical formulation for optimization, not evidence that selected configurations will generalize. The author advises calibrating targets through preliminary simulations and notes that the top-ranked design depends on the chosen objectives and constraints, so it may not be the most profitable candidate.

Key ideas

• Normalize each objective by a positive target before combining objectives with a weighted sum.
• Use weights to express the relative importance of normalized performance measures.
• Represent constraint violations as penalties that lower the optimizer's objective for infeasible designs.
• Choose targets and bounds after preliminary simulations, and avoid bounds that leave no feasible candidates.
• The optimizer's top-ranked design reflects the chosen formulation and need not maximize profit alone.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.