This study deals with efficiency assessment of probability function by using different time series. Eight hydrometric stations in Karun drainage basin selected and their maximum instantaneous discharge completed for an 85 years period. Tow sets of 10-10 and 5-5 established in each station by splitting stations' long record into series with different lengths of five and 10 years intervals that finally led to obtaining 192 time series. Frequency analysis of maximum annual instantaneous discharge studied in time series of two sets. Analysis consisted seven methods, Normal, 2PL Normal, 3PL Normal, 2P Gamma, Pearson type III, L Pearson type Ill, Gumbel type I distributions. Goodness of fit studied by calculating sum squares of relative deviations and chi-square. Compression of the Frequency Analysis' results indicates that estimated floods, up to 50 years return periods, are very close to each other and the differences increase in high return periods. Estimated floods with selected functions are close to calculated floods by empirical probability methods up to 100 years return periods. In spite of variation in period's covering time series with equal length, Tree parameter functions are most selected as best distributions. Results show that L Pearson type III has a good flexibility and selected as best fitted distribution in 192 time series of two sets. Results show a considerable Convergence from application of tree parameter distributions. To test the major obtained results, 72 hydrometric stations over the Zagross zone, covering 10 provinces, selected and flood frequency analysis carried out with different time series. Log person type III selected as best distribution in 53 stations. The tree parameter distributions in general, with 79 percent t-elative frequency, are the hest probably function in flood frequency analysis in region under consideration.