Probability and Statistics: What They Are and How They Work Together

Probability and statistics are two subjects that, in introductory high school math courses, almost always travel together, even though they look at uncertainty from opposite directions. As OpenStax puts it, statistics deals with the collection, analysis, interpretation, and presentation of data; probability, on the other hand, is the mathematical tool used to study randomness, meaning how likely an event is to happen. Someone rolling a die uses probability to predict; someone who has already logged a thousand rolls uses statistics to describe what actually happened.

Descriptive statistics and inferential statistics

Statistics splits into two branches. Descriptive statistics organizes and summarizes data that has already been collected: it counts, sorts, and calculates averages. Inferential statistics, according to OpenStax, is the set of formal methods for drawing conclusions from “good” data (collected with sound criteria), and it comes into play when observing an entire group isn’t possible. That requires two key concepts: the population, meaning the collection of people, things, or objects under study, and the sample, a portion of that population selected to gather information from. A value calculated over the whole population is called a parameter; the same kind of value calculated over a sample is called a sample statistic, used to estimate the corresponding parameter.

A clean example of descriptive statistics on its own is a population count: according to ISTAT, Italy’s national statistics institute, the resident population stood at 58,943,673 as of June 30, 2026. That’s not an estimate drawn from a sample: it’s a count, updated periodically, of the entire population — the simplest case of descriptive statistics applied to a real number.

The measures that summarize a group of numbers

To summarize a set of numbers, descriptive statistics relies on three measures of central tendency. The mean is the sum of all values divided by how many there are, the same addition-and-division operation covered in Arithmetic, just applied to a whole group of numbers instead of two. The median is the value sitting at the center of an ordered list: at least half the data is less than or equal to it, and at least half is greater than or equal to it. The mode, more simply, is the value that shows up most often in the set.

Worked example: OpenStax lists the ages of 20 students in a fifth-grade class: 9, 9.5, 9.5, 10, 10, 10, 10, 10.5, 10.5, 10.5, 10.5, 11, 11, 11, 11, 11, 11, 11.5, 11.5, 11.5. Adding up all the ages and dividing by 20 gives a mean of 10.53 years. Sorted, the middle value (the median) falls right inside the “10.5” cluster, very close to the mean — a sign that in this case the data is fairly symmetric, with no outliers pulling the mean off to one side.

When a set instead contains values much higher or lower than the rest, the mean and median start to drift apart. OpenStax ties this directly to the shape of the distribution: the mean is affected by outliers that don’t influence the median, so in a left-skewed distribution the mean tends to sit below the median, and in a right-skewed one it tends to sit above it. That’s why, when a single number has to describe a very uneven group, the median is often the more honest choice.

How spread out is the data: standard deviation

Knowing only the mean of a group isn’t enough: two classrooms can share the same average age with very different makeups, one where every age is close together, another mixing much younger and much older kids. Standard deviation is the number that captures exactly this: how far, on average, the values in a set sit from their own mean. It’s small when the data clusters near the mean, larger when the values are spread out.

In the 20-age example, the sample standard deviation OpenStax calculates is 0.72: a low figure, consistent with a group of children who are nearly all the same age. The formula uses the same letters and symbols found in expressions from Algebra: s = √[Σ(x − x̄)² / (n−1)], where x̄ is the mean and n is the number of values. One technical detail explains why the differences from the mean get squared instead of simply added up as they are: adding the deviations from the mean without squaring them always produces zero, because the positive and negative gaps cancel each other out.

Classical probability and frequentist probability

Probability, too, can be calculated in more than one way. According to the Stanford Encyclopedia of Philosophy, the classical interpretation defines the probability of an event as the fraction between the number of cases in which the event occurs and the total number of possible cases — a formulation that traces back to the mathematician Abraham de Moivre, who described it as the fraction with, in the numerator, the number of ways an event can happen, and in the denominator, the total number of ways it can either happen or fail.

Worked example: a six-sided die has six possible outcomes, all equally likely by the symmetry of the object; the set of these outcomes is called the sample space. The classical probability of rolling a 4 is therefore 1 favorable case out of 6 possible cases, or 1/6. The same logic applies to a coin: two possible outcomes, heads or tails, each with a probability of 1/2.

Classical probability works well when outcomes are equally likely by construction, like the faces of a fair die. But many real-world events don’t come with that kind of built-in symmetry: how many times a machine breaks down in a year can’t be worked out by dividing favorable cases by possible cases. That’s where frequentist probability comes in, defined by the Stanford Encyclopedia of Philosophy as the relative frequency with which an attribute actually shows up within a reference class: the event is observed many times, and the proportion of times it occurred is calculated. Further along in a math curriculum, once these observed frequencies turn into continuous distributions studied through functions and limits, the territory becomes calculus.

Independent events and the gambler’s fallacy

Two events are called independent when one occurring doesn’t change the probability that the other occurs. Statistics LibreTexts defines independence this way, and this exact property is what dismantles one of the most common mistakes tied to games of chance.

Worked example: after five coin flips in a row that all came up heads, many people think that on the sixth flip tails is somehow “more likely,” as if the coin needed to balance out the sequence. Statistics LibreTexts calls this reasoning the gambler’s fallacy: people who fall for it imagine a natural adjustment compensating a run of heads with more tails later on. But the coin carries no memory of the flips before it: on the sixth flip, the probability of heads is still 1/2, exactly as it was on the first.

There’s one related observation that tends to cause confusion and is worth separating from this fallacy: across a very large number of flips, the proportion of heads does tend to settle near 50%. That doesn’t mean past flips influence future ones, though: it simply means that, spread across enormous numbers of independent trials, any single “unusual” stretch (many heads in a row) counts for less and less in the final proportion. The distinction is subtle but precise: no individual flip “corrects” anything; it’s the average over a huge number of flips that settles down on its own.