A 30 year (1986-2015) daily discharge data of five gauging stations namely, Kantamal, Kesinga, Salebhata, Sundargarh and Tikarapara of middle reach of the Mahanadi river basin of Odisha, India were analyzed for predicting peak flood discharge at various return periods (probability of exceedance). Twelve probability distribution functions (PDFs) viz., Normal, Log-Normal, Pearson, Log-Pearson, Weibull, Generalized Pareto, Extreme Value Type III, Gumble-maximum, Gumble-minimum, Generalised Extreme Value, Exponential and Gamma were tested by comparing the values of chi-square test, and two statistical measures of error, namely, root mean square error (RMSE) and mean absolute error (MAE). The analyses were carried out using the FLOOD software. The Gamma distribution was found to be the best fitted PDF for the Kantamal, Kesinga, Sundargarh and Tikirapara gauging stations of the middle reach of Mahanadi River as these stations had the lowest values of the Chi-square, RMSE and MAE. However, Extreme Value Type III was the best PDF for the Salebhata station. The predicted peak daily discharge ranged from 11012.1 to 31728.1, 1917.9 to 16076.8, 574.8 to 5100.3, 328.9 to 5501.9 and 11092.3 to 31728.1 m3s-1 for the Kantamal, Kesinga, Salebhata, Sundargarh and Tikirapara, respectively for probability of exceedances ranging from 10 to 90%.