Zilli doesn’t know your future. It knows your answers — and what they add up to over ten years.
Two minutes on what the simulation actually computes, what it holds still, and how to read the numbers because of it.
What the simulation actually does
You answer a few questions about how you date: how many evenings a week you can give it, how many new people you meet, how often a first date leads to a second, how long you stay before you know. Zilli takes those numbers and plays them forward one day at a time for ten years — booking dates on a calendar that only holds so many, letting relationships start, run, and end — then it throws that decade away and lives the whole thing again, hundreds or thousands of times, because one decade tells you nothing about the range.
That is the entire trick. It is arithmetic on your assumptions, run all the way to the end. No part of it is a prediction about you.
It is worth doing because compounding is real and human intuition is bad at it. A conversion rate that feels fine at one date, and a wait that feels reasonable in one relationship, do something to ten years that nobody can work out in their head. That arithmetic is what Zilli is for. There is no second thing.
What the model assumes
Five simplifications do most of the work. Each one buys the model its speed and costs it something real — here is the trade, and what it means for what a result can say.
Every person the simulation meets is drawn the same.
There is no chemistry variable and no “the one” — the model can say how often things work out at the rate you gave it, never that it was going to be them.
Your answers hold still for all ten years.
Nobody in the model gets better at dating, gets busier, moves city, or ages, so year ten describes someone dating exactly the way you date today.
Each stage moves on steady odds.
Relationships end here by a weighted coin, not by the fight or the job offer or the good month that changed everything — the drama is real, it just isn’t modeled.
Everything it believes came from you.
There is no hidden national average underneath, so if a number looks wrong it is wrong — the review page holds every one of them, and every one is editable.
It models your process, never a person.
No simulated future says anything about the human across the table on Thursday.
Not all of it is invented: several of the engine’s default numbers — the window before a spark goes cold, the early-date conversions, the shape of the funnel — line up with published research on how real dating funnels behave. The citations live in the project’s research notes, not here.
How to read what comes back
The range is the product. The curve and the band around it are the answer; the midpoint is only the middle of it. If your decade reads “somewhere between three and seven years,” that spread is the finding, and rounding it to five throws the finding away.
Comparisons travel further than levels. What changes when you change one thing is the sturdiest number on the page: the model’s blind spots are the same in both runs, so they largely cancel when you set one against the other. “Two evenings a week instead of one pulls this in by about a year” survives a great deal of wrongness in the absolute figures.
Every number is prefixed “given my answers.” Read them that way and they are useful. Read them as an oracle and they are just numbers with decimal points in them.
What Zilli won’t do
And then change one
The assumptions are yours. Disagreeing with one and running it again is not a workaround — it is the whole point.
Came from your results? Your browser’s back button keeps them.