LOGIC & PUZZLES / CHANCE
Probability
without the mystery.
Chance is easier to discuss when the possible outcomes and the assumed process are written down.
THE MAIN IDEA
A probability belongs to a model of an experiment.
Probability describes uncertainty by assigning values from zero to one to events under a defined model. Zero means an event is impossible in the model; one means it is certain. A fair six-sided die gives each face probability one-sixth because the model assumes six equally likely outcomes. If the die is weighted or the table is tilted, that assumption may fail even though there are still six faces. The sample space is the set of possible outcomes, while an event is a group of outcomes that answers a question. For a fair die, “roll an even number” includes three faces, so its probability is three out of six. For two dice, the equally likely elementary outcomes are ordered pairs, not simply the eleven possible sums. Counting sums without accounting for how many pairs produce each sum creates misleading odds. Independent events do not change one another's probabilities; drawing without replacement from a deck usually does, because the first card changes what remains. A large sequence can also look streaky. Independent coin flips have no memory, so a run of heads does not make tails “due” on the next flip. Long-run frequency may approach a model's probability over repeated trials, but any finite sample fluctuates, especially when the sample is small. Randomness can produce visible patterns; the eye's desire for alternating outcomes is not a rule of chance. A probability about a simplified game does not automatically transfer to medical tests, weather, investments, or personal risk. Those require relevant data, evidence quality, and a model that captures how outcomes are generated. When communicating probability, distinguish “one in ten under these assumptions” from “guaranteed every ten tries.” No individual sequence is promised. Simulating many trials can reveal how distributions behave, but random-number software itself uses a particular process and does not establish the assumptions of a real event. Define the event, list outcomes, and state why they are considered equally likely before doing arithmetic. When the evidence comes from observed frequencies, record the sample and method. This turns probability from a vague label into a checkable argument while leaving space to update beliefs when better information appears.
TRY THIS
Write out the two-dice sample space.
List the 36 ordered pairs of faces for two fair dice. Count the pairs that total seven, then compare with totals of two and twelve. The totals are not equally likely even though the individual dice faces are, because different numbers of pairs create them.