Why Overlapping Events Break Naive Addition

If you roll a standard six-sided die, can you ever be certain what lands face up? Before the dice stops rolling, every number feels equally reachable and equally hidden. Yet, across hundreds of rolls, a strict pattern emerges from the uncertainty. We measure that balance on a single scale stretching from zero all the way to one.

A value of zero marks something impossible, like rolling a seven on that standard die. A value of one marks absolute certainty, like rolling a number between one and six. Everything else in reality sits somewhere between those two edges. Picture all six possible faces laid out inside a single fenced boundary.

This collection of every outcome that could happen is called the sample space. Suppose you place a bet that the die lands on an even number. That event carves out three winning spots, namely 2, 4, and six. To find the probability, compare the target area to the entire territory.

Three winning outcomes divided by six total outcomes simplifies down to 1/2. The formal rule follows directly from what your eyes already saw. We divide the count of favorable results by the total count of possibilities. Notice what happens to the faces you did not pick.

1 3 and five. Together, the winning events and the losing events blanket the entire space without gaps. Because the total territory must always sum to one, the leftovers are easy to find. The probability of missing is just one minus the probability of hitting.

Now, suppose your friend offers a new wager on the same single roll. You win if the roll is even or if the roll is greater than four. Your intuition whispers that you should calculate both chances and add them up. There are three even numbers and two numbers greater than four, suggesting five winners.

If we kept adding events together like that, the total could quickly exceed one. Here is why that naive addition fails on our six-sided die. Look closely at the face showing the number six. Six is an even number, but six is also strictly greater than four.

By counting both groups separately, we counted that single face twice. To repair the tally, we add both probabilities together and subtract the shared overlap. 3 / 6 + 2 over 6 - 1 / 6 gives 46. That fraction simplifies to 2/3, capturing every winning face without double counting.

Now imagine the die rolls under a cup so you cannot see the face. A friend peaks under the cup and reveals the number is greater than three. Suddenly faces 1 2 and three vanish from consideration entirely. Your entire world shrinks down to three possibilities four 5 and six.

Inside the smaller world, how likely is it that the number is still even? Only two numbers four and six satisfy being even inside that surviving group. So the probability changes from 1/2 up to 2/3. That ratio is conditional probability.

Measuring an event after information narrows the possibilities. What happens if you lift the cup and roll the die a second time? The first roll is finished, resting quietly on the table with no memory. Its outcome cannot push, pull, or nudge the second tumbling die in any direction.

Because neither roll influences the other, we call them independent events. To find the chance of rolling two consecutive sixes, we multiply their separate chances. 1/6th of the first possibilities branches into 1/6th of the second. Multiplying those fractions together shrinks the winning target to 1 in 36.

Every calculation returns to that original scale stretched between zero and one. Uncertainty is not the absence of order, but order waiting to be counted. When you slice possibility into equal pieces, the future becomes measurable. Thank you for exploring probability from scratch.

Continue with our previous video, how neural networks actually learn visualized in 3D. The link is in the description and top comment.