A lousy way to do memory prediction is “X happened in the past, therefore X will happen in the future.” It’s too based on specific circumstances. What you want is principles. You want mental models.