The aim is to construct a reliable risk evaluator in order to supply reliable information to the control systems or to alert the driver. When LTR is equal to zero, the vehicle has a stable roll dynamics. The risk becomes as high as the indicator goes towards +/-1. Both extreme values characterize wheel takeoff, but this does not always signify an accident. The occurrence of rollover is then represented by wheels takeoff without rollover, i.e. by a roll angle enough high. Sensibility analysis of the (driver-vehicle-infrastructure) system allows us to deduct the most influencing parameters on the risk. These parameters are represented by stochastic models. Their probabilistic laws are obtained by using maximum entropy principle based on available information about the uncertainties. From these laws, Monte Carlo simulations will be realized. Then, the analysis is made on the histogram of the LTR laws, the probability that LTR exceeds given thresholds, the time of threshold first crossing, and the laws of the LTR maximal values. Finally, we determine the critical threshold and the moment of the risk occurrence to feed the risk evaluator. The tests will be made on Prosper simulator of Sera-CD and on real scenarios at high rollover risk.