Darrieus turbines face difficulty to self-start, especially in environments with fluctuating inflows that cause them to deviate repeatedly from their designed operating parameters. To elucidate the self-starting process in this study, a three-bladed Darrieus rotor was simulated numerically with vector diagrams to facilitate visualizations on the rotor behaviors. Based on segments of the average rotor torque coefficients (Cτ), the self-starting process consisted of linear and accelerated phases, with the first two segments in the linear phase and the next two segments in the accelerated phase. The simulation showed that the self-starting process was largely influenced by dynamic stalls. The rotor experienced difficulty to self-start in the first segment as it encountered a region of “dead band” with a negative mean cyclical caused by a reverse dynamic stall. This dynamic stall and its corresponding dead band disappeared in the second segment, which initiated the transition into the accelerated phase. In the third segment, forward dynamic stalls that formed boosted the generation and accelerated the angular speed of the rotor toward its peak. Finally, without any dynamic stalls formed in the fourth segment due to reduced values of the inflow angles on the blades, they reduced drastically until the rotor reached its steady phase. Outcomes from this work demonstrate that understanding the effects of unsteady aerodynamics is vital to improving the self-starting process. Potential design improvements on the rotor that address this aspect include static and dynamic pitching, blade flaps, intracyclical control, and flow controls using blowing and suction mechanisms.