This paper presents a novel strategy for spherical search, initially proposed in 2019, which demonstrated promising performance but suffered from high computational costs due to matrix projection operations. The proposed strategy employs random projection to reduce computational demands while maintaining similar performance levels. Additionally, we introduce the integration of multiple local and global variance helpers to enhance overall performance and exploration capability. Our experimental results, conducted on a diverse set of benchmark problems from the CEC2014 competition, demonstrate that the proposed method preserves performance while significantly reducing computational complexity.