Preliminary to investigating the length of the climatic period whose average gives the best (minimum variance) estimate of the next year's value, previous studies are examined and the results of five are replotted onto a standard scale. All indicate that prediction one year ahead from an average based on only 20 years, or so, is better than one from a standard 'climatic normal' of 30 years. Monte Carlo simulation of the prediction process suggests that slight changes with time in the means, whether real or caused by instrumental or observational changes, in most climatic records reduce the record length for optimum prediction.