Dynamic Portfolio Optimization Using Evolution Strategy

● The classic Markowitz model that optimizes for the Sharpe ratio has proven to be suboptimal. ○ Summary is used directly as prediction. ○ Variance is not a good risk measurement since it penalizes positive shocks and says little about tail risks ● Other risk measurements such as Value-at-Risk and Expected Shortfall introduce non-linear, non-convex risk constraints and render the mean-variance approach less applicable ● Time series such as security prices often violate the assumption of independent sampling and homoscedasticity We thus devise an Evolution Strategy (ES) for the multi-period portfolio optimization, which dynamically adjusts the asset weights in hope for significantly higher returns. 1 PROFITABILITY OF ES