Answer

Hello.
Explanations after the fact are easy to produce. And that is precisely their weakness. Think, for example, of a computer that gives you a sequence of numbers: …3,5,7

Now you are asked to predict what the next number will be. There are two possible theories:
A. It is a sequence of odd numbers, and according to that the next number is 9.
B. It is a sequence of prime numbers, and according to that the next number is 11.
Which of the two is more correct? Both fully explain what we have seen so far. If the computer now outputs 11, that confirms theory B. Why? Because it succeeded in predicting a future result, and not merely in explaining past results.
A criterion that relies on the past would see the two explanations as equivalent, and therefore fit to the past is weaker evidence. The ability to predict a future result is a very powerful confirmation tool.
To put it differently: it is easy to find an explanation for results that are already known. That is an ad hoc fit, and you can always propose infinitely many explanations for any set of results. But it is harder to hit upon a future result in advance.
And one more formulation: in a theory that makes a prediction, you are taking a risk—you may be proven wrong. In a theory that only explains the past, there is no risk at all. If it works, it works, and that’s that. The greater the risk, the greater the payoff—that is, the more correct the theory is. A theory that is willing to put itself to the test of falsification is a theory that is more confident in itself and more correct.