method

this page says what the number is made of, so you can decide how much of it to believe. if you came here to find the hole in it, most of the holes are listed below, by us.

what the number is made of

one table. the probability of dying within a year, for every single year of age from 0 to 119, male and female, published by the US Social Security Administration, Office of the Chief Actuary.

• publisher: Social Security Administration, Office of the Chief Actuary
• table: period life table, single year of age, 0–119, by sex
• vintage: {table_year}
• basis: period — this matters, and it has its own section below
• population: the United States, and only the United States
• terminal age: 119. nothing is extrapolated past it
• licence: public domain, as a work of the US federal government
• retrieved by hand on {retrieved_date} · sha256 {file_sha256}
• source: ssa.gov/oact/STATS/table4c6.html

no second table is blended in. no modifier is applied: we do not ask about your smoking, your weight, your sleep, your blood pressure or your postcode, and we do not adjust the number for any of them. what you get is the plain population baseline, because the baseline is the only part we can defend.

it is a US dataset. if you are not in the US, the figure on the page is a fact about americans your age, and this sentence is the whole of our defence of it. we would rather say that than quietly hand you a number about somebody else's country.

it starts from the age you have already reached

most death calculators take life expectancy at birth and subtract your age. that is wrong, and it is wrong in one direction: too short. you have already survived every hazard that dragged the at-birth figure down.

in the SSA's 2022 table, life expectancy at birth for US men is 74.74 years. in that same table, a man who has already reached 65 has 17.3 more years in front of him — which puts him at about 82, not about 75. he did not gain anything. he simply stopped being averaged with the people who did not get to 65.

so the arithmetic here is conditional: it walks the table from your exact age forward, year by year, and asks how many of the people standing where you are now are still standing later. the number is about the remaining road, not the whole road.

it is a median, not a prediction

the headline is the median — the point where half the people the table counts as being like you are still here, and half are not.

which means, plainly: half of the people like you are still here past it. the figure is not a date, not a limit, and not a thing you can be behind on. the distribution has a long tail to the right, which is why we publish the median rather than the mean: only the median lets us write that previous sentence and have it be true.

the page also gives you the spread. one in four get well past the median. one in ten are gone well before it. nobody here knows which one you are.

the spread is the product

this is the part a death-date site cannot sell you, and it is the part that is actually true.

a single date is a more satisfying object and a worse one. it is precise where the data is not, it invites you to plan around a number that was never about you, and the only way to make it feel earned is to stop mentioning how wide the uncertainty is. we show the band by default, at the same moment as the figure, not behind a tap on the word "about".

the band is wide. it is still narrow enough to notice that you are not going to be here indefinitely, which is the only thing we are asking you to notice. you do not need a precise number for that. you have never needed a precise number for that.

the table is probably too pessimistic about you, if you are young

a period table takes one year's death rates — the rates measured across every age last year — and assumes they hold, unchanged, forever. they have not held unchanged for a century. they have improved, decade after decade.

so for a young person, a period table applies the death rates of today's 70-year-olds to a person who will be 70 in forty years, and that is very likely too harsh. a period table systematically understates how long the young live. the direction is certain. the size of it is not, which is why there is no correction factor here and no number in this paragraph.

notice which way that cuts. the error flatters our pitch: it makes the figure smaller and the page more alarming than the truth probably warrants. we are telling you anyway, because a product about honesty that keeps its most convenient error quiet is just a product.

it drifts later every day you survive

because the estimate is conditional and recomputed from your exact age, the median moves. every day you are still here, the table hands a little of it back.

the consequence is strange and true: you cannot outlive this estimate. it is not a countdown that empties. it is a line that keeps re-drawing itself slightly further out, the way a horizon does. this is the single most counterintuitive thing the arithmetic does, and the reason there is no ticking clock anywhere on this site.

the arithmetic runs in your browser

the table is about twelve kilobytes. the engine is about five hundred lines. both are loaded into the page you are reading and run on your own device, in under a millisecond, offline if you like.

your answers are never sent anywhere. there is no form submission, no API call on the path to the number, no analytics event carrying your inputs, and no server that has ever held your birth year or your mother's. this is not a policy we are promising to keep. it is the shape of the thing: there is nowhere for the data to go, because the page never speaks to us at all.

and because the inputs live in the URL rather than in an account, you can bookmark your own result and have it forever, with no login, including after we are gone.

what it does not know, and what we refuse to ask

it does not know your genes, your blood pressure, your driving, your doctor, your luck or your last scan. it does not know whether you are ill. we do not ask.

we also refuse some inputs on purpose:

• race and ethnicity. complete US life tables by race exist, and the gaps in them are large and real. those tables measure healthcare access, exposure and racism at a population level — not the body of the person holding the phone. handing someone a shorter clock and calling it their data would be the worst thing this site could do. the honest cost: it makes the figure less accurate for some people, and we would rather say that than fix it this way.
• anything about your mental health. no history, no diagnosis, no mood, no inferred state, no score, ever. those route to support, never to arithmetic.
• a prognosis. if you are seriously ill, the arithmetic is removed, not refined. we do not ask how long you have.
• the wellness shelf. supplements, fasting windows, cold plunges, sauna protocols. there is no defensible individual-level effect on all-cause mortality for any of them, and no price at which we would sell you one.

and what we do with the answer: nothing. the estimate is never transmitted to, sold to, or made queryable by an insurer, a lender or an employer. no advertising for life insurance, funeral services, burial plots or annuities is ever sold against it. there is no API anybody can query by identity, because there is no record to query. the rest of that list is at never.

check it

the engine and the vendored table are public. the table has a published hash, and a test that refuses to display any figure from a file that does not match SSA's own published life expectancies. if the file is wrong, the page shows you nothing rather than something plausible.

nobody here knows when you are going to die

not the table, not the arithmetic, not us. the table knows the shape of a population. you are one person, and it has never met you.

it is going to be wrong about you specifically. it is right about the shape of the thing. that is the whole method.