The SO algorithm has been shown to be effective in solving global optimization problems and engineering applications by emulating the behavior of snake swarms. However, traditional SO algorithms exhibit a high degree of dependence on empirical control parameters when selecting thresholds, making the algorithm prone to local optima or limited convergence speed when faced with highly nonlinear optimization problems. The Multi-Directional Yielding Snake Optimizer (MySO) introduces a novel approach by proposing a pull-push optimization strategy, which involves the dynamic exploration of the search space through the application of push and pull forces. This strategy aims to broaden the search, avoid falling into local optima, and improve the global search capability of the algorithm. Furthermore, the Sobol Sequence and the Subtraction Average-Based Optimizer (SABO) were introduced to achieve diversity in population initialization and to enhance information interaction and dynamic adaptation between individuals. MySO is compared to six MAs, SO and its variants on the CEC2020 benchmark function, and eight engineering problems. The final experimental results show that MySO exhibits significant performance advantages in global optimization problems.