How NameYear Makes Sure Every Puzzle Is Solvable

Every NameYear puzzle shows a “class roster” of six first names that were all common among American babies born in the same year. You slide to the year you think they were born. The obvious worry with a game like this is fairness: what if the names simply don’t point to the answer? This article explains how each roster is checked before it is ever used.

Step 1: what each name says about a year

The data is the U.S. Social Security Administration’s baby-name file: for every year since 1880, how many babies of each sex were given each name (names given to at least five babies). For any name and year we can compute its share: the fraction of that year’s babies who got the name. Jennifer’s share is large in the early 1970s and tiny in the 1930s; Mary’s is the reverse.

Step 2: combine six names into one guess

Suppose all six children were born in some year Y. The chance of drawing exactly these six names from that year’s babies is, roughly, the product of their six shares in Y. Do that for every year from 1925 to 2025, and you get a score for every candidate year: years where all six names were common at once score high, years where even one name was rare score low. Normalized so the scores add up to one, this is a probability for each birth year (statisticians call it a posterior with a flat prior).

From that curve we take two numbers: the single most likely year, and the narrowest range of years that holds 80% of the probability.

Step 3: throw out unfair rosters

A roster is only used if the data’s most likely year is within two years of the true answer, and if its 80% range is narrow: between 10 and 18 years wide depending on the round. Rosters that fail are discarded and the generator draws new names. So every puzzle is, by construction, solvable from the names alone by someone who knows the data perfectly.

A worked example

Here is a roster from a past puzzle (round 3 on 2026-10-20):

Carla · Damien · Tasha · Clayton · Marvin · Natasha

The data’s best guess for this roster is 1980, and 80% of the probability falls between 1973 and 1985. The answer was 1979. A player who guessed the data’s best year would have scored 905 of 1000 points.

Rounds and scoring

The five daily rounds draw from different parts of the popularity list, from the top-10 names of the year in round one down to names ranked in the hundreds in round five, so the puzzle gets harder as you go. Each round scores round(1000·e^−error/10): a perfect guess earns 1000, five years off earns 607, ten years off 368. Asking for two extra names caps the round at 800.

After each round the results page shows every name’s popularity curve, so you can see why the answer was what it was. The same probability model powers the Guess my group tool, which estimates when a group of friends was born from their first names.

Source: U.S. Social Security Administration national baby-name files, 1880–2025. Puzzles are generated in advance and published as one file per day.