Verbal memory

What the game is doing behind the word

Words appear one at a time from a pool of 123 concrete English nouns. For each one you answer a single question: has this word already appeared in the current run? A correct answer scores one point. A wrong answer in either direction, calling a repeat new or a new word seen, costs a life, and the game tells you which kind of mistake you made. Three lost lives ends the run, and your score is your total of correct answers, saved locally as your best.

The draw is not uniform. Once you have seen more than three words, each new card has a 40 percent chance of being a repeat pulled from your seen list, and a 60 percent chance of being a fresh word from the remaining pool. That ratio is why mashing "New" collapses quickly: it is right at most 60 percent of the time and burns three lives in short order.

How to run deep

This is recognition, not recall. You never have to reproduce a word, only tag it as familiar, and recognition is dramatically stronger than people expect, which is why scores here dwarf digit-span numbers. The failure mode is not weak memory, it is weak encoding: a word you swiped past without attention leaves no trace to recognize later. So slow down for half a second on every new word and do something with it. Picture the object, or note something odd about the word itself. "Zeppelin" costs you nothing; the run-killers are the mild furniture words you barely registered.

A subtlety worth knowing: even when you answer wrong, the word still counts as seen from then on. If you lose a life calling "quartz" new when it was a repeat, or seen when it was fresh, the correct answer next time it appears is "Seen" either way. The word the game just corrected you on is the one word you know its status for with certainty, so bank it.

When genuinely unsure, lean "Seen" in long runs. Deep into a game most of the pool is behind you, so repeats outnumber fresh words among the cards you would hesitate on, and a familiar-feeling word usually feels that way for a reason.

What a good score looks like

Runs of 30 to 60 are typical. Past 100 you are doing well, and at that depth you have necessarily seen most of the pool, so the game shifts from memory test to attention test: one lapse while a stray thought passes is how strong runs end. Because focus is the binding constraint, your score varies more with sleep and distraction than with anything resembling verbal talent, and comparing scores across a noisy afternoon and a quiet morning is not a fair fight.

Honest design notes

The pool is finite at 123 words, all common nouns, no proper names or trick spellings. A long enough run touches all of them, after which every word is a repeat; the game keeps dealing rather than declaring victory, so extremely deep runs gradually become a test of whether you noticed the pool ran dry. There is no timer on your answer, no leaderboard, and no server: the words, your streak, and your best score all live in your own browser tab. One quirk to watch: the repeat draw does not exclude the word you just answered, so occasionally the same word appears twice back to back. It is not a glitch, and the answer the second time is "Seen".