Classical mathematical routines as Markowitz portfolio or Black-Litterman portfolio are must-have instruments for a portfolio manager who relies on quantitative analysis. However, these tools are not flexible enough when it comes to custom target selection, accounting for risks or constraints that are hard to optimize for with well-known tools. They aim to maximise fixed target tied to a fixed risk source (as Sharpe ratio) with fixed market beliefs (assets covariance and some predictive factors). And all that an investor can do - just click the "optimise" button and analyse the backtest without having any real possibility to influence the observed drawbacks.
We offered to our client a more generalised view on the portfolio optimisation and asset allocation routine. We gathered the requirements about desired goals: Sharpe, Sortino, Calmar ratios, associated risks: Maximal Drawdown, CVaR, and necessary constraints: market neutrality, L2 regularisation on the weights, etc. Also, we have added an option to rebalance the portfolio instead of finding one single set of weights for the allocation of the capital on the assets. Then, we have developed an optimisation model based on the evolutionary algorithms that could find the optimal rebalance scenario and associated allocation weight for each of the assets. This procedure can be repeated with different options for obtaining another optimal portfolio.