This paper presents a method for a short-term stock index prediction. The source data comes from the German Stock Exchange (being the target market) and two other markets (Tokyo Stock Exchange and New York Stock Exchange) together with EUR/USD and USD/JPY exchange rates. Neural networks supported by a genetic algorithm (GA) are used as the prediction engine. Except for promising numerical results attained by the system the special focus in the paper is on the problem of elimination of dead chromosomes, i.e. the ones which cannot be properly assessed.