The performance of the Tryfos MMSE estimator and the mean of the Bayesian posterior are comparable, both strictly better than the constant baseline (except for the 1000 meters event, where the observed record saw no change during the hold-out period). The largest difference between the two predictors was observed for the Marathon event, where the posterior mean estimator performed significantly better than both Tryfos MMSE and the constant baseline. While the magnitude of the difference is partly driven by the scale of the marathon times, in Figure 5 we see that the posterior median estimator correctly tracks the actual observed records.