Auto-correlation functions (ACF, grey and black) and GP kernels (QP blue, SHO green, QPC red) with largest fitted likelihoods of all RV data sets sorted by spot configuration (from top to bottom ONE, POL, RAN, and TWO). The SHO kernel has a less prominent correlation decay and puts more weight on correlations at larger τ in comparison to the QP and QPC kernels. We reiterate the fact that the hyperparameters of the kernels are specifically sensitive tothe correlations at shorter time-lags because of the linear decrease of the number of data-point pairs N_τ with time-lag τ, i.e. .