How can we evaluate the probability of things happening based on historical data?

Question

I try to evaluate the probability of things happening in the next period based on how many times they happened in the previous period.

For example, in the first period, case A happens one time, case B happens two times, and case C happens three times. Can we evaluate the probability of the above three cases happening in the second period?

Answer 1

Fundamentally, this is the only way to accurately evaluate the (frequentist) probability of something happening. Most of machine learning, forecasting, and insurance is predicated on being able to measure probabilities from historical behavior.

The core assumption is the same as science itself -- that there exists stable patterns and order in our world.