The loss of load probability (LOLP) is an appealing means used in operational reliability and capacity planning studies to evaluate the installed reserve margins and incentivize generation capacity investments in electrical power systems (EPSs) since it can express reliability explicitly and visually. LOLP recommendations can significantly differ depending on how the stochastic properties (variability and uncertainty) of the load are represented in the analysis. The stochastic load properties are represented by a continuous random process. This approach's difficulty with LOLP evaluation problems creates a huge computational burden. A simplified approach discretizes the load stochastic process into random variables (RVs) indexed by time. However, there is still a fundamental question: on which time intervals are RVs considered, hourly, daily, weekly, monthly, or seasonally? and based on which stochastic properties are considered, peak or average load variation in the supposed time interval? Therefore, this paper examines different approaches to modeling and quantifying the load's stochastic properties and their impact on the calculated LOLP values. Moreover, analytical expressions are derived to incorporate them into EPS’s LOLP assessment efficiently. Besides, the approximation errors of the analytical LOLP expressions are discussed.