The Markowitz portfolio allocation is one of the most well-known portfolio-allocation approaches in modern algorithmic/quantitative trading and is based on modern portfolio theory. The idea here is to take the co-variance between the returns of all the trading strategies in our portfolio and account for that when allocating risk to individual trading strategies to minimize portfolio variance while maximizing portfolio returns. It is a convex optimization problem and has many well-known and well-understood techniques to solve. For a given level of portfolio variance, it can find the best allocation scheme to maximize portfolio returns by building what is known as an efficient frontier curve, which is the curve of optimal allocations for the trading strategies in the portfolio for different levels of risk. From there, as our risk appetite grows or shrinks, and as more strategy results are available as more trading days are seen, it is straightforward to rebalance the portfolio by using the readjusted efficient frontier.

For Markowitz allocation, we can state the following:

• Allocation seeks to maximize diversity of the different trading strategies in the portfolio, by ensuring that strategies with uncorrelated returns have risk allocated to them.
• While in other allocation methods, the risk allocation for strategies that have poor performance would have dropped close to 0, here even losing strategies have some allocation assigned to them. This is because the periods in which these losing strategies make money offsets periods where the rest of the portfolio loses money, thus minimizing overall portfolio variance.