Who was Enrico Fermi?
Enrico Fermi was an Italian-American physicist, Nobel Prize winner, and one of the defining scientific figures of the twentieth century. His name became associated with estimation questions because he was known for breaking big, fuzzy problems into smaller quantities that could be approximated.
What makes a Fermi problem different?
A Fermi problem asks for a useful estimate when exact data is missing, delayed, or impractical. You decompose the question into parts, estimate each part, keep the units straight, and check whether the result feels plausible.
Because the target is an order of magnitude rather than false precision, these articles normally use no more than two significant digits and write most calculations in scientific notation; both practices are introduced in the Fermi reference.
Why news-based Fermi problems?
Every question here starts with a real news article and a specific claim or number. From that, we build a rough calculation that forces two tests: do you have useful numeric pegs in memory, and do you understand the world well enough to combine those pegs into a decent model?
News often tells us that a number is large, small, shocking, wasteful, impressive, or historic. Fermi problems ask the next question: compared with what? A public cost, a climate claim, a market trend, or a consumer habit becomes more meaningful when placed next to a scale we can understand.
Why this catches bad assumptions
The arithmetic can be simple while the assumptions are revealing. If two readers differ by a factor of 100, the interesting question may be whether they assumed different populations, prices, time periods, densities, conversion factors, or physical constraints.
Then we compare the estimate with authoritative references. There is a real chance you may disagree with the sourced answer or with my interpretation of it. I am no expert. If you disagree, comment and explain why.
How the Calibration Score works
Earn points for accuracy across four areas: pegs, model, math, and final result. Higher is better: 100 is best and 0 is worst. The full-of-it image shows the inverse percentage: 100 minus the Calibration Score.
| Area | Maximum points | Standard |
|---|---|---|
| Accurate pegs | 30 | Within 25% of the reference values earns 30 points, within 50% earns 20, within 100% earns 10, and more than 100% wrong earns 0. |
| Sound model | 30 | A reliable calculation earns 30 points. A model that sort of works earns 15. A model that cannot answer the question earns 0. |
| Correct math | 10 | No math mistakes earns 10 points. Mistakes with no major impact earn 5. Mistakes that substantially change the answer earn 0. |
| Accurate result | 30 | Within 10% earns 30 points, within one order of magnitude earns 20, within two orders earns 10, and farther off earns 0. |