I think the point was more that it is comparatively easy to develop your new theory to fit known 'facts', so that testing against them should usually be regarded only as a sanity check rather than as additional evidence. To be a bit pithy, prediction is worth more than postdiction.

(Of course, if your theory fails to comport with known 'facts' but successfully predicts the result of new observations, then that is deeply exciting!)

Here's another way to explain it:

Since string theory is extremely tunable, you can just overfit it to all the data we already have about how the universe works.

So the only way to get some certainty would be to have it make a prediction that we don't know the answer to, and then test it. At this point in time, we don't have the means to conduct such an experiment.

> you are using the word fact li ke it could somehow still be false.

Quite so: whether my sister's name is "Andrea" is a fact. It might or might not be true that's her name.

Perhaps I'm using the word "fact" incorrectly, but I think it's correct: for example, whether the earth is flat or curved is a fact; the proposition that it is flat is false. But the word "proposition" seems rather abstract, in a context like this.

In the context of physics (of which I'm not an expert), I believe you are using the term 'facts incorrectly. What you are calling a fact, in physics would be a prediction produced by our existing models, that is supported by observation.

Our current model of gravity predicts what happens when an apple falls, and that's what we observe. That is what you're calling a 'fact'.

String theory can produce all of these same predictions. If it only predicted the same stuff we can already predict, then it's useless. If it can make a new prediction, then its potentially valuable. So if it can make a definitive statement that will be true or false that says when you do X, Y will happen, and right now we can't predict what will happen after X, or our current model says Z will happen then we're very close to adding value. But Y could still be wrong.

So finally if string theory saying X will cause Y, our current theory doesn't know what will happen after X, or thing something other than Y will happen we can run and experiment and falsify one of them. If we observe Y happens then great, string theory is useful in modeling our universe. If anything else happens, then it hasn't provided modeling/predictive value.

The current complaint against string theory is it's not about to make a solid prediction and let us run any test to see that it's true or false. So essentially if it's useless.

It like a theory that says an alligator is ornery because it has a lot of teeth and no toothbrush. That's falsifiable. If I give it a tooth brush then it should stop being so angry. If it one and it stops my theory is right if I give it one and its still angry my theory is wrong.

Instead imagine a theory. When asked "why are alligators so angry?" Answers "because i said so". Well that's not very useful. You can use that to "explain" everything, but predict nothing. That's why falsifiable predictions are the test of value for a model/theory.