Fair enough. An explanation is in order, then. Apologies for length.
Suppose you have a hypothesis about some aspect of the world, and you want to know if it's true. So, you try to design a way to test your hypothesis. A test of your hypothesis would be something that, if it has a certain result, means your hypothesis is more likely to be true, and if it has another result, means your hypothesis is less likely to be true¹.
Now, how do you know how to test? Well, you look at your hypothesis, and ask yourself what it would predict in a certain situation. If you think people named John are more likely to prefer the colour green than the average person, then your hypothesis would predict (for example) that if you survey lots of people, you'd see that the number of people named John who prefer green would be greater than the number of people named Ann who prefer green. So, your test for your hypothesis could be to make a survey. This is only one of the possible predictions you could make, and only one of the possible tests, but if you make many different tests and your hypothesis passes them all, each time it's more likely that it is, in fact, true. You are never completely, 100% no mistakes possible sure, but you get closer to that every time.
To test a hypothesis, you need to know what it predicts in a given situation. For some situations, your hypothesis might make no particular predictions, for example the John hypothesis above would tell you nothing about the populations of crocodiles in Egypt. It simply has nothing to do with it. But, and this is important, your hypothesis needs to make predictions for at least some situations. If I say that I have a monkey sitting on my left shoulder, then you can make plenty of predictions about that, including some really simple like "If I look at your left shoulder now, I'll see a monkey". But suppose I say the monkey is invisible, and intangible, and never makes a sound, and similarly for every possible way to test for the monkey. Then the monkey hypothesis is unfalsifiable. It simply doesn't make any difference whether the monkey is there or not. The universe looks exactly the same in either case.
As knowledge, the imperceptible monkey hypothesis is useless. There's no possible way to know if it's true or false since I cannot test it, and even if it was true, it would tell me nothing useful, I don't really know any more than I used to about the world. That's the basic idea of why hypotheses need to be falsifiable.
Now, you might argue that your flood hypothesis is not untestable, and that's partially true. Specifically, you accept "Floods will feature in many old legends" as an acceptable test of the hypothesis, but nothing else. That basically means that, once you've observed the legends, your hypothesis makes no further predictions about the world. That's equivalent to arguing that, other than creating legends, the Great Flood did absolutely nothing of any relevance. If it had, that would be a possible test, look for the marks of what the Flood did in geological formations and the like. But, if you accept no further evidence, then you are arguing for a Flood that covers the entire world, kills most of its population, and its only perceptible effect is creating stories. Do you see why that is incongruous?
¹Karl Popper is the guy who is most associated with the general idea, which is called falsificationism. The version I'm describing here has some differences with the one Popper proposed, for example, he said that results either made a hypothesis false or fail to make it false, rather than less likely or more likely.
What is not the answer to this question?