Abstract Quantitative prediction of gas hydrate formation risk is critical for the successful implementation of risk-based approaches to hydrate management in subsea production. Here we use a high pressure, stirred, automated lag-time apparatus and a high pressure acoustic levitator to experimentally obtain smooth probability distributions describing stochastic hydrate formation. Robust, repeatable hydrate nucleation and growth rate probability distributions as a function of induction time and subcooling are measured for various gases, shear rates and inhibitor dosages in systems with interfacial areas ranging from (0.1 to 60) cm2. The results reveal that new engineering models can be used to reliably predict hydrate formation probability. Importantly these new models have solid theoretical foundations which enables them to be generalized with confidence to industrial systems.