Understanding crop yield responses to rainfall is essential for food systems adaptation under climate change. While there are ample evidences of crop yield responses to seasonal rainfall variation, the geographic sensitivities and driving mechanisms of sub-seasonal rainfall events remain elusive. We used long-term nationwide observations to explore the sensitivity of maize and soybean yields in response to event-based rainfall across Chinese agroecological regions. While maize and soybean yield showed concave downward responses to event-based rainfall depth at the national scale, these responses were differed considerably among regions. These differences can be primarily explained by soil moisture preceding rainfall events, soil erosion and sunshine hour reduction during rainfall. Our projections reveal that focusing on seasonal rainfall or national-level sensitivity analysis suggests a 0.3–5.9% increase in maize yields due to future rainfall, yet considering spatial variations unveils a contrasting reality, with maize yields declining by 9.1 ± 0.3% under a medium-range emission scenario (SSP2–4.5) by the end of century (2085–2100). The future rainfall effect on soybean yield is the opposite, leading to a 20.6 ± 3.9% reduction nationally without spatial consideration, but an increase (by 7.0 ± 1.0%) when spatial variations are factored in. These findings underscore the critical necessity of incorporating regional variation in yield responses to sub-seasonal rainfall events, which could otherwise lead to vastly different impact estimates, even reversing the expected crop yield response to future rainfall change.