We derive a Bayesian model to forecast the continuation of a cumulative progression of records. We show that the model compares favourably to past least-square prediction models for records in running events. We also check the validity of the model for the case where we have access to the underlying attempts, using data from athletic events in the Olympics.

We present a Bayesian approach for modeling a time series for a cumulative record that takes the form of the maximum or minimum of a sequence of attempts, in the absence of data for the underlying attempts. We discuss the derivation of the likelihood function, sampling of the posterior via PyMC3, and forecasting the distribution of future records.