We repeat this same calculation for each of the remaining events in the Tryfos dataset, but with only 5000 posterior samples per chain since we observe little difference in our results compared with the longer chains. To compare the performance of the Bayesian posterior with the MMSE, we compute the MSE of both the posterior mean and the MMSE estimator compared to the actual observed records over the forecasted period (1983-1997). We also include in our comparison a constant baseline estimator that always predicts the last observed record.