A school surveys 200 students. Of the 80 who play a sport, 50 also play an instrument; of the 120 who do not play a sport, 40 play an instrument, so 90 students play an instrument in all. Pick a student at random and P(instrument) = 90/200 = 0.45. Now you learn that the student plays a sport. The 120 non-athletes are no longer possible, so the denominator becomes 80, and P(instrument | sport) = 50/80 = 0.625. The condition did not change the overlap of 50; it changed which group you divide by.
The formula says the same thing with probabilities: P(A∣B)=P(B)P(A∩B)=80/20050/200=8050. Dividing by P(B) rescales B so that it becomes the whole sample space. Swap the roles and you divide by a different group: P(sport | instrument) = 50/90 ≈ 0.56, because 90 students play an instrument. Same overlap, different denominator, different answer.
Not every question with two events is conditional. P(sport and instrument) = 50/200 = 0.25 asks about the whole school: what share of all 200 students do both. P(instrument | sport) = 0.625 asks about athletes only. The giveaway is the reference group. If the question narrows attention to a subgroup, such as "of the athletes" or "on rainy days", it is conditional, and that subgroup's size goes in the denominator.
Think of it like this. Zoom a camera onto one section of a stadium: once you know the person is in the student section, you count only the people in that frame and ask what share of them wear red. Zooming changes the group you divide by, not the people in the overlap.
Core idea. Knowing that B happened replaces the original sample space with B itself, so the chance of A becomes the share of B's outcomes that are also in A: P(A∣B)=P(B)P(A∩B), which usually differs from P(B∣A).