In this research the level of accuracy for Chow regression and stochastic methods is compared for estimating annual peak flood in central Alborz region. The annual peak flood data in this region were incomplete, and regression method was used for data completion. 23 gauge stations with 20 years common data were selected for the analysis. The observed data at each station was divided into time series of 10, 15 and 20 years preparing 414 time series of annual peak floods for the analysis. Using 7 important frequency distributions including, Normal, 2 parameters Log Normal, 3 parameters Log Normal, 2 parameters Gama, Pearson type 3, Log Pearson type 3, and Gumbel 1the probabilities were accounted for each of these time series. Then the best distribution was chosen and the annual peak floods for 2, 5, 10, 15, 20, 25, 30, 50, 100, 500, and 1000 years return periods were estimated. Peak discharges for these return periods were estimated using chow’s regression and Stochastic methods, and then compared using probability indices such as, MSE and MBE. This research showed that Chow’s regression method gives better results for estimating annual peak flood in the central Alborz region.