Draw a card from a standard 52-card deck and peek only at its color: red. Your world has just shrunk from 52 cards to the 26 red ones. Of those 26, 13 are hearts, so P(heart∣red)=2613=21. Before you peeked it was 5213=41. The information did not change the deck; it changed which cards are still possible, and so it changed the denominator.
A table of counts makes this concrete. In a class of 30, 12 students play a sport, 8 play music and 5 do both. Given that a student plays music, the chance they also play a sport is 85: look only at the 8 musicians. Given that a student plays a sport, the chance they play music is 125. Same overlap of 5, different denominators, which is why P(A∣B) and P(B∣A) usually differ.
The joint probability P(A and B) looks similar but has no 'given': it is the overlap measured against everyone, 305 for sport and music. The formula connects the two: P(sport∣music)=8/305/30=85. When knowing B leaves the chance of A unchanged, P(A∣B)=P(A) and the events are independent; here 85 is far from P(sport)=3012, so they are not.
Think of it like this. Conditioning is a search filter. Filter a deck to red cards and 26 remain; hearts are half of what is on screen, so P(heart∣red)=21. Filter to hearts instead and every result is red, so P(red∣heart)=1. Swapping the filter changes the answer.
Core idea. P(A∣B) is the chance of A once you restrict attention to only the cases where B already happened.