Wind speeds of various average return periods are estimated from the series of hourly mean wind speeds assumed independent. This hypothesis of independence is suggested by characteristics of the largest value distribution of Gaussian and non-Gaussian series. The confidence in this estimator of wind speed is subsequently examined when the shape of the instantaneous distribution of the hourly mean wind speed series is available but its parameters are unknown. It is found that the estimators are generally accurate even when the available record is short, e.g. 1 to 5 years of hourly readings. This conclusion is supported by test cases involving records of hourly mean wind speeds in Sydney and Lisbon. On the other hand, the conventional method based on Gumbel lines fitted to yearly maximum wind speeds is inaccurate when based on records of similar sizm.