Thread:TheHumanAmbassador/@comment-26907577-20190924151043/@comment-26907577-20190926131839

While I understand your point, I will propose the following:

Predictive power is, in essence, the ability of your hypothesis to accurately tell the outcome of an event before the event. Correct?

We do not have any experiments we can perform to gain new data to test our hypotheses on.

However, predictive power is useful primarily for one thing: showing off. While it is cool for a hypothesis to hold up to new evidence, it doesn't actually tell us anything. If we get new evidence that contradicts our hypothesis, we must revise it. If we get new evidence that supports our hypothesis, great.

In either case, we did not need new evidence to test our hypothesis.

So the point of induction in this regard is not to create testable hypotheses, but working and probable ones. That is, given the evidence in favor of a hypothesis and lack of disproving deductive evidence, it is more probable to be true than the null hypothesis. (What would you set the p-value at for game theories? It's irrelevant as we can't calculate these things very well.)

Interestingly, however, we have very few null hypotheses. This is an issue.