Several concepts and analysis techniques originating from Nonlinear Dynamics have been applied to electroencephalography recordings of epilepsy patients to predict seizures. An early prediction of an upcoming seizure would dramatically increase the therapeutic possibilities for this common neurological disease. We suggest standards to assess seizure prediction performance of time series analysis techniques. We present assessment of three methods originating from Nonlinear Dynamics with respect to their ability in predicting epileptic seizures. © 2004 American Institute of Physics