Probability describes how likely an event is. When you learn that likelihood by carrying out an experiment, you use experimental probability. The result is evidence from what happened, not a promise about what must happen next.
Turn results into a proportion
A trial is one repetition of the experiment. A success is the outcome you are counting; it need not be a winning outcome. If a spinner lands on blue 18 times in 60 spins, its experimental probability of blue is
The denominator includes every spin, not just the spins of other colors. You can express the same result as or 30%. These forms all describe the observed share of blue outcomes.
Theoretical probability comes from a model, such as equal sections on a fair spinner. Experimental probability comes from observations. They can differ because of chance, or because the model does not describe the real spinner well.
Example: predict a future count
Use the 60-spin result to predict the number of blue outcomes in 200 more spins. Multiply the new number of trials by the observed proportion:
Predict about 60 blue outcomes. This does not mean every group of 200 spins will contain exactly 60. It is a reasonable planning estimate if the spinner and spinning method remain the same.
A useful check is scale: 200 is a little more than three times 60, so the predicted blue count should be a little more than three times 18. Sixty fits that expectation.
Example: combine experiments using counts
One class observes 12 blue outcomes in 30 spins. Another observes 18 blue outcomes in 70 spins. The combined experimental probability is
Do not simply average the two classes' percentages. The second class supplied more trials and therefore more observations. Adding successes and adding trials gives each individual trial equal weight.
From these combined results, a prediction for 150 new spins is blue outcomes.
More trials help, but do not create certainty
A fair coin can land heads seven times in ten tosses. That short result does not establish that heads has probability 70%. With many independent tosses under unchanged conditions, relative frequencies tend to settle nearer the true probability.
Still, a coin is not “due” for tails after several heads. For a fair coin with independent tosses, the next toss has probability of heads regardless of the earlier sequence. Long-run behavior is not a rule forcing short-run compensation.
Try it yourself
A player makes 42 of 70 practice shots. Estimate how many of 120 similar shots the player will make. Would 72 successful shots be guaranteed?
Answer: The observed proportion is so the prediction is It is not guaranteed. Changes in conditions and ordinary chance can change the actual count.
What to remember
Experimental probability is successes divided by trials. Predict a count by multiplying that proportion by the new trial count, and describe the result as an estimate. Combine unequal experiments by adding counts, not averaging their percentages.