The paper calculates three drawdown expectancy functions: for a system with positive (u > 0), neutral (u == 0), and negative profit expectancy (u < 0). The essential formula is (7). You can see in fig. 1 how the Q functions, which are proportional to the drawdown expectancy, develop over time.

We're using the second curve in fig. 1, which is a square root function, for normalizing the simulation DDmax to a 3 years period. This means when a one-year simulation produces a drawdown X, the normalized drawdown would be Y = X * sqrt(3).