Optical coherence tomography angiography (OCTA) offers unparalleled capabilities for non-invasive detection of vessels. However, the lack of accurate models for light-tissue interaction in OCTA jeopardizes the development of the techniques to further extract quantitative information from the measurements. In this manuscript, we propose a Monte Carlo (MC)-based simulation method to precisely describe the signal formation of OCTA based on the fundamental theory of light-tissue interactions. A dynamic particle-fixed model is developed to depict the spatial-temporal behaviors of the tissue phantom: the particles are initialized and fixed in specific locations with wavelength-dependent scattering cross-sections and are allowed to travel over time. We then employ a full-spectrum MC engine to faithfully simulate the formation of OCT and OCTA images. A simulation on a vessel-mimicking phantom demonstrated that speckle characteristics in OCT as well as decorrelation maps in OCTA could be successfully reproduced. We further illustrate the usefulness of our method on the quantitative OCTA by extending it to simulate the gradual saturation of decorrelation in OCTA-based velocimetry. We believe our method will serve as a valuable tool for studying OCTA theory and inspire better solutions and metrics for non-invasive flow velocity measurement.