There is a need to supply coastal managers with statistically defensible hazard predictions that can be used to implement coastal setbacks and other management policies. The goal of this article is to evaluate the widely used single-transect method, as well as several new methods: t-binning, IC-binning, polynomial methods, and eigenbeaches, to identify which method(s) best predicts a 50-year eroded shoreline position. The polynomial and eigenbeach methods allow for acceleration (the rates vary with time). The methods are compared using data from nine beaches on Maui, Hawaii, and four sets of synthetic data. Evaluations of the methods are based on an information criterion, color maps of residuals, long-term (50 year) predictions, and cross-validating the most recent shoreline, which has a short-term span of 5–9 years. The newer methods identified significant rates at 74% of the transects, vs. 0% for single-transect on beaches in Maui, Hawaii. The cross-validation results showed that the polynomial and eigenbeach methods, without acceleration, best predicted the most recent shoreline. Contrary to the cross-validation results, synthetic results showed that the polynomial and eigenbeach methods with acceleration predicted the 50-year shoreline better than methods without acceleration. Nonacceleration methods predicted short-term positions better, and acceleration methods predicted long-term positions better. We conclude that the polynomial and eigenbeach methods improve the significance of the rates compared with the single-transect method.