Statistically modeling them using the prior dates of leap seconds alone isn't a good method (if that's what you did, not claiming that it is). Knowledge of the

Statistically modeling them using the prior dates of leap seconds alone isn't a good method (if that's what you did, not claiming that it is). Knowledge of the past leap seconds doesn't really tell you anything about when future ones will be. A better way would be to model the difference between UTC and UT1 and model the rate of change there. Probably the best would be to model the earth's rotation rate and factor in the likelihood of the events that cause variances. One major source of those events is earthquakes.