We present a comprehensive treatment of the pixel-lensing theory and apply it to lensing experiments and their results toward M31. Using distribution functions for the distances, velocities, masses, and luminosities of stars, we derive lensing event rates as a function of the event observables. In contrast to the microlensing regime, in the pixel-lensing regime (crowded or unresolved sources) the observables are the maximum excess flux of the source above a background and the full width at half-maximum (FWHM) time of the event. To calculate lensing event distribution functions depending on these observables for the specific case of M31, we use data from the literature to construct a model of M31, reproducing consistently photometry, kinematics, and stellar population. We predict the halo- and self-lensing event rates for bulge and disk stars in M31 and treat events with and without finite source signatures separately. We use the M31 photon noise profile and obtain the event rates as a function of position, field of view, and S/N threshold at maximum magnification. We calculate the expected rates for WeCAPP and for a potential Advanced Camera for Surveys (ACS) lensing campaign. The detection of two events with a peak signal-to-noise ratio larger than 10 and a timescale larger than 1 day in the WeCAPP 2000/2001 data is in good agreement with our theoretical calculations. We investigate the luminosity function of lensed stars for noise characteristics of WeCAPP and ACS. For the pixel-lensing regime, we derive the probability distribution for the lens masses in M31 as a function of the FWHM timescale, flux excess, and color, including the errors of these observables.