Scientists have shown how to combine the forecasts of a collection of suboptimal ‘delay embedding’ predictors for time series data. This work may help improve the forecasting of floods, stock market gyrations, spatio-temporal brain dynamics, and ecological resource fluctuations.