We can predict some probabilities theoretically. For example, the theoretical chance of getting a 6 on any one throw of a fair six-sided die is 1/6. In theory, if we throw the die six times, we should expect to see one 6 shown. In reality, this often doesn't happen.
We can calculate the relative frequency (experimental probability) once we have collected some real-life data. To calculate relative frequency for a particular event, divide the frequency of that event by the total number of trials.
e.g. If you threw a die 50 times, and got 12 sixes, the relative frequency of a six would be 12/50, which simplifies to 6/25. You could also convert this to a percentage - 24%.