A good prediction includes not only a date on which to judge the prediction, making it falsifiable, but also a subjective probability, to account for genuine random chance. After all, there are few things in life we can be 100% certain about. For example, you may think the polls show such overwhelming support for a candidate that they will definitely win an upcoming election, but there is always a slight chance that they die of a heart attack, or be scandalously revealed as taking bribes from the opposition to be their puppet.

Here’s an example of a well-formatted prediction, from https://slatestarcodex.com/predictions-bets/: “Gay marriage will remain legal throughout a Trump presidency [confidence: 95%]” (prediction made 2016-11-15). That page also has some more detail on the theory of making predictions.

For this situation, david_w could give a prediction like “I think there is a 90% probability the EFF will support banning strong encryption by 2021-07-01.” Then, if the prediction turns out to be false, it either means David was wrong and he should have been much less confident, or he was right and encountered that unlucky one-in-ten unusual situation. And you, the observer who would personally estimate a different probability of that prediction coming true, could calculate the probabilities of those two situations being the case by using various formulas, including Bayes’ theorem. (David himself could also calculate the probabilities of those two situations, and use that to decide whether to change any of his views on the EFF.)