Study shows that a novel machine-learning approach using global climatic patterns can improve seasonal prediction of harmful algal blooms.