Earthquake source parameters, such as slip on the planes of finite faults, have spatially variable distributions, and seismic ground motions significantly fluctuate in accordance with the characteristics of complexity. The complexity of fault parameters should thus be evaluated properly. Although self‐similar models consider the complexity of the fault parameters, they have not been exhaustively used to validate earthquakes that occur in Japan. In this study, for multiple earthquakes in Japan, the slip complexities of finite faults are evaluated using the self‐similar model. Slip spectra are fitted using the von Karman autocorrelation function; furthermore, the Hurst exponents and correlation lengths are compared with previous studies dealing with earthquakes around the world. Differences between source types are also examined. The Hurst exponents H are unaffected by magnitude, and the difference between the source types is small. The median value of H is 0.77, which is nearly the same as the global average value indicated by Mai and Beroza (2002). The correlation lengths along‐strike have positive correlations with magnitude; furthermore, no differences were observed between source types. The characteristics are nearly the same as the global average value; however, the correlation lengths down‐dip correlates positively with magnitude. Nevertheless, there are differences for each source type. The characteristics have saturating trends for each source type, which is attributed to the thickness of the seismogenic layer. The moment magnitude saturates at 6.3 and 8.4 for crustal and interplate earthquakes, respectively. This is nearly consistent with the boundary between the first and second stages of three‐stage scaling.