Why some simulated lives end at four
2026-11-05
The most common complaint about a statistically honest life simulator is that it is broken: 「I got a life that ended at four. That can't be right.」
It can. In the countries with the highest child mortality, more than one birth in ten does not reach age five. A simulator that samples from those countries' life tables will produce short lives at roughly that rate, whether or not it feels appropriate.
How the draw actually works
The age at death is not a random number between 0 and 100. It is sampled from the country's life table, split by sex — a table of the probability of dying in each year of life, given that you survived the previous one.
That distinction matters. A uniform random number would give you a 1% chance of dying at any age, which is nothing like reality. A life table gives you a high probability of death in infancy in some countries, a very low one in childhood almost everywhere, and a rising probability from midlife onward.
| Country | Deaths before 5, per 1,000 | Share reaching age 5 |
|---|---|---|
| Nigeria | 115.6 | 88.4% |
| Niger | 110.7 | 88.9% |
| Somalia | 101.1 | 89.9% |
| Japan | 2.4 | 99.8% |
| Estonia | 2.0 | 99.8% |
The gap between the top and bottom of that table is a factor of 57. A run that ends at four is not an outlier in the global distribution — it is the ordinary outcome of being born in the wrong place, which is exactly the thing the simulation is trying to show.
Why not soften it
There are three obvious ways to make early deaths less jarring, and all three are worse:
- Re-roll early deaths until the life reaches adulthood. This makes every country look the same and destroys the only thing the model is measuring.
- Cap the minimum age. This is the same lie with a floor, and it produces the strangest result of all: a country with high child mortality that never shows a child dying.
- Remove children from the simulation entirely and start at 18. This is defensible for a classroom, and it is why a teacher mode exists — but as a default it hides the single largest component of the global life-expectancy gap.
What the model is and is not
It is one draw from a simplified life table. It is not a prediction about you, and it is not health advice. The same country drawn twice gives two different lives, and neither is a forecast.
What it is good for is scale: draw a thousand lives from one country and the pattern that emerges is the country's actual mortality profile. Draw one and you get a story. Both are the point.
More