Experimental probability is probability based on past experience, that is, by calculating a probability by examining results of a reasonable number of previous trials or experiments. This type of probability is used usually when there is no mathematical way of logically determining a probability. (A mathematical way, for example, would be to calculate the probability of getting heads when tossing a fair coin. In this case, experimental probability does not apply. The probability is 0.5, because it's not dependent on previous tosses of the coin.)
An example of experimental probability is the probability that a particular student will fail an upcoming mathematics test. By looking at a lot of that student's previous test results, an experimental probability can be deduced and applied to predict the likelihood of that student passing the upcoming test.