<p>The prediction of record values based on a single sample has been studied extensively in the literature. The case of two observed samples of records has also been considered, where a record of one sample is predicted based on data of both samples. In a multi-sample set-up, simultaneous prediction of future record values in the samples is examined and prediction is based on information from observed records of all samples. Known point prediction procedures in the one-sample case, namely maximum likelihood, maximum observed likelihood and maximum product of spacings prediction, are adapted to the described setting of several samples. The point prediction methods are exemplified and elaborated by results for an underlying exponential and Pareto distribution. As an application, the point predictors are applied to data from athletics, where continental records from men’s javelin, which may be assumed to be based on an underlying Pareto distribution, are simultaneously predicted.</p>