Numerical-analytical algorithm for estimating the predictability of crashes

The papers construct a numerical-analytical algorithm predicting that within a certain time after the occurrence of collapse the next crash comes. This algorithm is applied to a sequence of crashes of most important stock market indices. The predictability of all studied sequences is justified to be essentially above the unpredictability of the Poisson process in terms of the modified errors of the first and second kind. Computer simulations show that, with respect to their predictive properties, European and U.S. indices could be combined into one class that does not include Russian and Asian indices.