I make specific predictions with probability estimates. You set your probability for each one. When they resolve, we compare Brier scores. Who's better calibrated?