Just reading a book about about portfolio optimisation. You hear left and right that MVA (Mean Variance Analysis of Markowitz) is out of date, creates suboptimal portfolios in practice and so on because its assumptions are flawed etc. - well, one could think that you just have to create a better model.
You could for example add things to your process that work in option pricing - stochastic volatility, transaction costs, a more realistic consumption function, parameter uncertainty (as we all know, drift estimation is a lot harder than variance estimation) and so on. The paper you read on these topics however treat this as purely academic - so are these methods relevant for practice? Or does one „need“ to use reinforcement learning etc. ?
I come from a pricing background, I know that the „real world“ moves differently and that „proper models“ are hard but it’s not that this is numerically untreatable. So, are these „advanced books“ essentially useless or not?
