Randomly partitions time series segments into train, development, and test sets; Trains multiple models optimizing parameters for development set, final cross-validation in test set; Calculates model’s annualized return, improvement from buy/hold, percent profitable trades, profit factor, max drawdown