Universality of algorithmic prediction for extremes of time series

Based on the precursor applied for prediction of strong earthquakes the paper constructs the algorithm forecasting when extreme events occur in sand-pile models. These seismic models, due to permanent loading of stress and rare instantaneous stress release exhibiting scale invariance, explain self-organized criticality of seismic process on a model level. The constructed algorithm is shown to be equally effective for activation and anti-activation scenarios of large model events. Modified type I and type II errors are applied as a measure of efficiency.