Abstract We develop a criterion based on a brute-force algorithm to systematically determine optimal fitting regions for fluctuation functions in Detrended Fluctuation Analysis (DFA) and Multifractal Detrended Fluctuation Analysis (MF-DFA). We analyze and compare results with several artificially generated time series with known parameters as illustrative examples of this technique. We show that crossover time scales can also be objectively and efficiently estimated using the introduced algorithm. Finally, we employ the proposed methodology to study the scaling behavior of a natural time series: monthly sunspot data previously filtered with empirical mode decomposition (EMD); in this case we find results comparable with those found in the literature.