Floods arise from the interaction between near-event precipitation and antecedent soil moisture, which regulates how efficiently precipitation becomes runoff. Studies often classify floods using catchment-level statistics and treat soil moisture as a bulk proxy, obscuring its regulatory role and the influence of specific soil layers on individual floods. This research utilizes explainable AI tools to quantify event-scale roles of precipitation and multi-layer soil moisture in flood generation globally. We trained a catchment-shared LightGBM model on the Caravan–GRDC dataset, predicting daily streamflow percentiles for 1,385 snow-free catchments. To isolate immediate forcing from antecedent state, we utilized lagged inputs: t−1 for precipitation and t−2 for soil moisture at four depths (0–289 cm). For 38,317 annual-maximum floods, we applied SHAP to decompose each prediction into predictor contributions, classifying events as precipitation-dominant or soil-moisture-dominant based on the largest absolute SHAP value. Results show that soil moisture is the predominant global flood driver, though precipitation dominance increases at the highest streamflow percentiles, indicating that the largest peaks require intense forcing to overcome storage constraints. The regimes exhibit distinct dynamics: Soil-moisture-dominant floods evolve slowly, with longer rising and recession limbs, regulated by shallow subsurface moisture (7–100 cm), and typically occur in larger, flatter, lower catchments with shallow water tables. Precipitation-dominant floods are flashier, with sharp rising limbs, stronger sensitivity to the surface moisture (0–7 cm), and more prevalent in smaller, steeper, high-relief catchments, with deeper water tables. Timing predictions reflect the challenge of capturing short-fuse storm dynamics relative to slowly evolving storage, with soil-moisture-dominant peaks well-timed, while precipitation-dominant peaks exhibited delays. This framework provides a scalable, event-based quantification highlighting catchment control, with soil moisture acting as an active regulator. This classification supports process-aware forecasting and climate-change adaptation by tracking regime shifts and indicating whether predictability depends on storm forcing or antecedent state.