This is a function of statistics, which is the basis of all scientific research. Let me give you a simple example.

Lets say I hypothesize that if I drop a ball it will fall straight down and hit the ground. I have two possible outcomes:

1. It hits the ground
2. It doesn't hit the ground

Assuming there is nothing in the way, and there are no tricks to this test, I will calculate the probability that the next drop will hit the ground given previous data.

Lets say I drop it 5 times. The probability that the next drop will hit the ground based on the data set is simply n/N; where n is the number of times it has hit the ground, and N is the number of total tests. In this case it has hit the ground 5 times and there are 6 tests (the five previous and the one I am about to do).

This gives a calculated probability of 83.3% that the sixth test will hit the ground.

Now lets say I do it 30 times, and I want to test the 31st. Again, the same two outcomes are possible within my test:

n/N = 30/31 = 96.7%

Now 100 times, testing 101st:

n/N = 100/101 = 99.0%

As you see, the first test did not give me much confidence that the ball will always fall down and hit the ground, but the 101st gave me a great deal of confidence that all future tests will produce the same result.

This is a trivial example, and the statistics for most tests are more complicated, but it shows the kind of effort that must go into science to get results that produce models that are likely to represent reality. Without sufficient data, no reasonable conclusion can be derived.