In studies, agents trained through self-play achieved winning rates of up to 95-96% against in-game AI.
Researchers have successfully used Double Deep-Q Networks (DDQN) to train AI agents. piater. - Street Fight
The ability to analyze opponent behavior, manage space, and adapt to changing situations, often referred to as "footsies". and adapt to changing situations
Beyond AI research, studies on human players suggest that "depth" in fighting games comes from: piater. - Street Fight
For a complete analysis of competitive Street Fighter, practitioners often refer to specialized wikis and community forums.