Some time ago, I read Aaron Brown's article in Wilmott magazine Jan-11, The Education of a Quant. Locked into a single risk model? Better no model at all.
He especially discusses the Markowitz' portfolio theory versus Kelly's approach of calculating the optimal fraction of wealth in a bet. Markowitz assumed people dislike risk on the ground of utility theory, ... this treats risk as cost ... Kelly said there was an optimal amount of risk.
Both recommending a portfolio composition, Markowitz and Kelly have their simplifications, but both approaches provide valuable insight.
The article also takes evolution as a possible model for understanding capital budgeting.
With my interest in evolutionary programming, I agree that the paradigm of evolution should be able compute optimal composition of a portfolio based on the principle of selection towards a fit function (with constraints or scenarios). The transparency of the methodology (evolution steps) will also provide valuable insight.
This all came into my mind, when we discussed the long term development plans for valuation and risk analytics in UnRisk recently. It is also true that different models, and the choice of backwards in time or forward solvers (PDE or SDE) have strengths and weaknesses. But they provide valuable insight - consequently we offer multi-model and multi-method approaches.
We also strongly believe in multi-strategy and multi-method approaches to understand risk aspects better.