Regeneration of reservoir models is successfully implemented with history-matched facies-probability map in multiple-point simulation. The facies-probability map is built from the average of selected models, which have similar well performance with observed data. The proposed method significantly reduces the number of forward simulations by a distance-based method. The final models preserve the provided well data and bimodal distribution and properly depict channel connectivity and facies ratio in the reference field. The final models predict reliable water-cuts and oil production rates for both pre-existing production wells and a newly drilled production well.