Maximum accuracy in the classification of subjects (genotypes) into normal and inverted populations as function of inversion age. The figure shows mean classification accuracy across t
values for all the simulation cases (2250). Our method achieves high accuracy for large inversions and a wide range of frequencies, as simulated by invertFREGENE. We find low accuracy in a discontinuous cluster of older and high frequency inversions, suggesting a critical behavior in the simulations at this point which we discuss further in the main text.