Frequency (or more correctly relative frequency) could be considered an experimental measurement of probability. Over a small number of trials the relative frequence measurments may depart considerably from the theoretical probabilty, but over a large number of trials the relative frequence should tend toward the theoretical probabilty.
For example in a coin tossing experiment the theoretical probabilty of throwing a "head" is 0.5 if the coin is fair. If we throw the coin ten times then on average we'd expect to see heads occur five times, giving a relative frequence for heads equal to 5/10, the same as the theoretical probability.
Of course with only 10 tosses there is also a good chance that we'd see other than 5 heads and the relative frequence differ considerably from the probabilty. But if we tossed the coin say one thousand times then we'd expect the relative frequency to be much closer to the theroetical value.