Results of experiment in Sect. 5.1: We compare the output of the PCA-based PSF subtraction approach with two versions of the half-sibling regression-based algorithm on four different data sets obtained with VLT/NACO. The images show the signal estimates in units of flux (i.e., in the same units as the input data). The numbers that label the planets are the respective −log₁₀(FPF) scores. The sub- and superscripted values refer to the uncertainties due to the placement of the reference positions used for computing the FPF. We find that for all planets, the HSR-based method achieves a higher −log₁₀(FPF) score than PCA. □