Fermi questions with answers, and how to solve them

Fermi questions with answers, and how to solve them

Fermi questions with answers look like trivia, yet none of them needs a formula. Break the question into two or three numbers you already know, multiply, and round to a power of ten. Getting the power of ten right is the answer. The digits after it are a bonus.

Most people treat "how many litres fit in an Olympic pool?" as a fact they either know or don't. So they pull a number out of the air, or give up. The question never asked for memory. It asks for a chain, and the chain is usually shorter than it looks.

A Fermi answer is a chain of things you already know

The questions carry the name of Enrico Fermi, the physicist known for good approximate calculations with little or no data. At the Trinity test in 1945 he dropped scraps of paper and watched how far the blast pushed them. From that distance he put the yield at 10 kilotons of TNT. The accepted figure is 21.

That miss is the benchmark. Off by two is a good Fermi answer. Off by ten means one link in the chain was wrong. Four worked examples:

QuestionChainAnswer
Litres in an Olympic pool50 m × 25 m × 2 m = 2,500 m³, and 1 m³ holds 1,000 litres2.5 million
Heartbeats in 80 years70 a minute × 60 × 24 × 365 × 80about 2.9 billion
People of 1.70 m stacked to the tip of the Eiffel Tower330 m ÷ 1.70 mabout 194
Seconds in a year60 × 60 × 24 × 365about 31.5 million

Each answer rests on numbers that fit on a napkin. The pool follows the World Aquatics size of 50 by 25 metres at a 2 metre minimum depth. The heart rate sits inside the normal adult resting range of 60 to 100 beats a minute. The tower is 330 metres to its tip.

Rounding hard is what keeps the chain honest

Every link carries some error. Take 60 or 80 heartbeats a minute instead of 70, and the lifetime total moves by about a seventh. Errors in different links also pull against each other: one comes out high, the next low. The Fermi problem article puts it plainly, overestimates and underestimates "help cancel each other out". A chain of rough numbers beats one bold guess for that reason.

So round on purpose. Call a year 30 million seconds. Call a person 2 metres if it makes the division easy. Settle 10, 100 or 1,000 before any digit. The How many? page uses the same trick with your own height: ten of you stacked reach 17 metres, so you count in columns of ten instead of single bodies. The Statue of Liberty measures 46 metres from heel to torch, not quite three columns, so the answer sits near 30. The real count is 27.

A rough answer can look like a wrong answer

The fair objection: 2.9 billion heartbeats is a model, not a fact, and a model can be wrong. True. A Fermi answer never claims the digits. It claims the size, and it shows its working, so anyone can see which link to fix.

A number pulled from the air shows nothing. When the real count turns up, the chain tells you where the miss came from. A lucky guess never does. That is also why a good Fermi game reveals the working after the answer, not just the number.

Fermi questions need one trusted anchor

Without a number you trust at the start, the chain floats. A question like how many thoughts a person has in a day has no measured answer to check against, so it trains confidence, not judgement.

Practise on questions that end with a real figure. Fermi sets three a day with a calculator built in and scores each guess as a multiple of the real answer. How likely? does the same for real odds, and Amountle asks how many things landed in a glass you can see. More dailies with a number at the end are in this list of games like Wordle but not words.

Five rounds a day, each with the real number and the working behind it.

Play today's How many?
Sources 5