Most retirement calculators ask for a balance, a yearly withdrawal and an expected return, and give back a line that slopes gently to an age. The line is the problem. Markets do not deliver an average return every year, and the order in which the good and bad years arrive decides more than the average does.
Take two people with the same savings, the same spending and the same average return over thirty years. One meets a bad market in the first five years, the other in the last five. The first is selling assets at low prices to pay for groceries, and those shares are gone when the recovery comes. The second barely notices. Same average, different outcome. This is sequence-of-returns risk, and a single straight line cannot show it.
So I built a simulator that does not draw one line. It draws thousands. Each run takes a plausible sequence of yearly returns, built from historical data by asset class going back to 1871, and plays a full retirement through it: withdrawals, taxes, the lot. At the end you do not get an answer. You get a distribution: in what share of those histories did the money last, and in the ones where it did not, when did it run out?
Three things turned out to matter more than I expected when I started.
Taxes change the withdrawal. A dollar from a taxable account, a traditional retirement account and a Roth are three different dollars. Which one you draw first, and in which year, moves the result by more than most arguments about asset allocation.
Spending is not flat. A fixed, inflation-adjusted withdrawal is a modeling convenience. Real people cut back after a bad year. A plan that allows even a modest cut when the portfolio falls survives many histories that a rigid plan does not.
The failures are the useful part. A 90 percent success rate sounds comfortable until you look at the other 10 percent and see that they all begin the same way: a poor first decade. That tells you where to hold a reserve and when flexibility is worth the most.
The tool runs on my own machine, in a container, on my own numbers. I built it because I wanted to interrogate the plan rather than be handed a verdict. It does not predict anything. It shows how wide the range of outcomes is, which is the one thing a single line with a confident slope hides.
None of this is financial advice. It is how I check my own arithmetic.