Fatigue damage due to random stress histories is typically estimated by either Rainflow range counting (a time history approach), or by the Rayleigh approximation (a power spectral density approach). These methods are not accurate in all applications. A new time history method of estimating fatigue life for a time history with undefined power spectral density is proposed. This is called the peak moment Rayleigh estimate, and it is based on the standard deviation of the random signal from knowledge of the odd moments of the peaks and valleys of the time history. This study also evaluates a recent spectral method of fatigue life estimation, the Lutes Single Moment estimate, using data from a previous experimental study by Sarkani for loadings with defined power spectral densities. This estimate is similar in form and ease of application to the Rayleigh estimate, but results in more accurate fatigue life predictions for processes with very wide bandwidths.