The time evolution of the electromagnetic emission from blazars, in particular high frequency peaked sources (HBLs), displays irregular activity not yet understood. In this work we report a methodology capable of characterizing the time behavior of these variable objects. The Maximum Likelihood Blocks (MLBs) is a model-independent estimator which sub-divides the light curve into time blocks, whose length and amplitude are compatible with states of constant emission rate of the observed source. The MLBs yields the statistical significance in the rate variations and strongly suppresses the noise fluctuations in the light curves. We apply the MLBs for the first time on the long term X-ray light curves (RXTE/ASM) of Mkn~421,Mkn~501, 1ES 1959+650 and 1ES 2155-304, which consist of more than 10 years of observational data (1996-2007). Using the MLBs interpretation of RXTE/ASM data, the integrated time flux distribution is determined for each single source considered. We identify in these distributions the characteristic level as well as the flaring states of the blazars. All the distributions show a significant component at negative flux values, most probably caused by an uncertainty in the background subtraction and by intrinsic fluctuations of RXTE/ASM. This effect interests in particular short time observations. In order to quantify the probability that the intrinsic fluctuations give rise to a false identification of a flare, we study a population of very faint sources and their integrated time flux distribution. We determine duty cycle or fraction of time a source spent in the flaring state of the source Mkn~421, Mkn~501, 1ES 1959+650 and 1ES 2155-304. Moreover, we study the random coincidences between flares and generic sporadic events such as high energy neutrinos or flares in other wavelengths. Comment: Accepted to A&A