Range and Spread

Two classes sit the same test. Both average exactly 70%. On paper they look identical — but walk into the two rooms and they could not feel more different. In Class A almost everyone scored in the high sixties or low seventies: a calm, bunched-up crowd. In Class B half the students scored in the 40s and half in the 90s: a room split down the middle, with hardly anyone actually near 70.

The average is the same. The experience is not. A single number for the centre has quietly hidden the thing a teacher would most want to know — how spread out the scores are. Very often the spread is what actually matters: it is the difference between a class that needs one lesson and a class that needs two completely different ones.

The mean, median and mode all answer "where is the centre?". Spread answers the other half of the question: "how stretched-out is the data?". This page opens the spread story with its simplest character — the range.

The range: crude but quick

The simplest measure of spread is the range: the gap between the largest and smallest values.

\text{range} = \max - \min.

It is instant to compute and easy to explain, which is why it turns up everywhere from weather reports ("today's temperatures ranged from 8° to 21°") to exam feedback. But it is the crudest measure of spread there is, because it depends on only two values — the two most extreme ones. Everything in the middle, however it is arranged, has no say at all. That fragility is why statisticians soon reach for sturdier tools.

Worked example 1 — a range in three seconds

Five friends' shoe sizes: 6, 8, 7, 11, 7. To find the range, spot the largest and the smallest — everything else is a distraction.

\max = 11, \quad \min = 6, \quad \text{range} = 11 - 6 = 5.

Done. Notice we never touched the three 7-and-8 values in the middle — the range simply doesn't look at them. That is its charm and its flaw in one.

Worked example 2 — same mean, very different range

Back to our two classes. Here are five representative scores from each, both chosen to average exactly 70:

Identical centres, and a range of 8 against a range of 53 — more than six times wider. The mean alone would have told you these classes were the same. The range shouts that they are worlds apart. This is exactly why a good summary reports a measure of spread alongside a measure of centre; one without the other is half a picture.

Worked example 3 — one outlier blows it up

Take a tidy, tightly-clustered data set: 20, 21, 22, 22, 23, 24. Its range is 24 - 20 = 4 — small, honestly reflecting how bunched the values are.

Now let one reading go wrong — a typo, a broken sensor, one unusual person — and slip a single 99 into the set: 20, 21, 22, 22, 23, 24, 99. The range leaps to 99 - 20 = 79. It has grown almost twentyfold, yet six of the seven values never moved. The bulk of the data is still tightly clustered around 22 — but the range now reports a wildly stretched-out set. That single point has hijacked the whole measure.

This is precisely the weakness that motivates better tools: the interquartile range, which throws away the extremes and measures the middle 50%, and the standard deviation, which uses every value's distance from the mean. Both are coming — and both exist largely to fix what the range gets wrong.

Watch one outlier stretch the range

Drag the orange point — the largest value — outward. The bracket spans from the minimum to the maximum, and the range grows with it, even though every other point stays put. One stray value, and the range balloons.

Same range, opposite shapes

Because the range sees only the endpoints, two data sets can have the identical range and yet be arranged completely differently. Both sets below run from 1 to 99 — range 98 each — but one clusters in the middle and the other splits to the ends:

Worked example 4 — reading a range in the wild

A weather station logs the midday temperature every day for a week: 14, 15, 13, 16, 15, 14, 27 °C. Six mild days and one freak scorcher.

\text{range} = 27 - 13 = 14\text{ °C.}

A range of 14° makes the week sound wildly variable — yet six of the seven days sat quietly between 13° and 16°, a spread of just 3°. The lone 27° has doubled the story. If you reported "temperatures ranged over 14° this week" you would be technically correct and deeply misleading. A better summary would flag the hot day as an outlier and describe the calm middle separately — which is precisely the instinct that leads to the median, the quartiles and the IQR.

The lesson threading through all four examples is the same: the range is a useful first glance — quick, honest about the extremes — but it should never be your only measure of spread, because it is entirely at the mercy of the two endpoints.

The range is built from exactly two numbers — the max and the min — and is blind to every other value in the data. That means two data sets with utterly different shapes can share the same range and fool anyone reading it alone. Compare:

Same range of 98, yet the first set is mostly tightly bunched and the second is violently divided. The range cannot tell them apart, because it never looked at the middle four values in either. This is exactly why statisticians prefer measures that use more of the data — the variance, the standard deviation, and the IQR all read the whole set, not just its two endpoints. Never report the range as if it describes the shape; it only describes the reach.

Averages get all the attention, but in the real world the spread is often the number that keeps people awake at night.

Two investment funds can advertise the same average return — say 7% a year. Fund A delivers a steady 6–8% every year; Fund B swings from −20% to +35%. Same average, wildly different risk, and risk is spread. The nervous investor and the thrill-seeker want opposite funds, and the average alone can't tell them apart — which is why finance is obsessed with a spread measure called volatility.

A bridge engineer does not design for the average lorry — the average is harmless. She designs for the worst case: the heaviest load the bridge might ever meet, out at the very top of the range. In safety engineering the extremes and the spread are the whole game, and the average is almost beside the point. Whenever the tails can hurt you — money, medicine, materials — spread beats average.

If you love an extreme range, visit Siberia. The town of Verkhoyansk holds one of the largest temperature ranges ever recorded in an inhabited place: from around −68 °C in the depths of winter to +37 °C in high summer — a range of well over 100 °C in a single location. Compare that with a tropical island, where the temperature might drift only 5° or 6° between January and July.

Same planet, same "temperature" variable — but the spread tells you almost everything about what it is like to live there. It also shows the range doing its honest job: as a headline number for "how far apart can the extremes get?", it is genuinely informative. The trouble only starts when people mistake that reach for a description of a typical day. A typical day in Verkhoyansk isn't 100° from anything — it's just cold.

See it explained