Annual series of maximum instantaneous flood peaks from 26 Pennsylvanian watersheds smaller than 200 square miles were analyzed by the Gumbel, log Gumbel, and log Pearson 3 methods. An empirical appraisal of how well each of these mathematical curves fits the data shows the general applicability of the straight Gumbel line. Consistent overestimation of long return period extremes results from using the log Gumbel analysis. Advantages of slight flattening for very rare expectations, which can result from the log Pearson 3 and the types of data sets preventing this from happening, are discussed. Corresponding histories of daily and clock hourly extreme rainfall annual series were compared to flood series for 24 watersheds that had more than one half of their drainage areas within a 14‐mile radius of the rain gage. Extreme rains that occur during the spring are more strongly linked to annual floods than yearly maximum rainfalls are. No usable relationship could be found between the extreme value statistics of rainfall and floods.