A Monte Carlo retirement simulation runs your figures over and over, each run drawing its own sequence of investment returns and inflation instead of assuming one smooth average. The score it hands back is the share of those runs that still had money left at the end. Ours runs 1,200 of them.
That number describes the model, not your life. A 90% result is not a promise about you, and a 60% result is not a prediction either. Both are statements about how a particular set of assumptions behaved when the order of good and bad years was shuffled.
This page is where each of those assumptions is written down, along with what the simulation refuses to touch. The tool itself builds your full Retirement Plan. If a figure it shows cannot be traced back to a line here, treat it as a defect and tell us.
There is no threshold that means safe. A score moves when the assumptions move, so the same household can score one way under one set of return and inflation assumptions and quite differently under another without a single thing changing in the household itself. That is why the assumptions below are published rather than buried: a score is only worth as much as the inputs you can inspect, and a simulation that will not show you its inputs is giving you a number, not an answer.
The more useful reading is the direction and the size of the gap. A score that moves a long way when one variable changes is telling you something about that variable. The absolute level, treated as a grade, is the part that misleads.
| Assumption | Value | Source |
|---|---|---|
| Investment return | 6.18% a year before inflation, 12.76% volatility | J.P. Morgan Asset Management's 2026 Long-Term Capital Market Assumptions, a 75/25 blend of US large-cap stocks and US aggregate bonds at the study's own correlation. Independently published, which is the point: the return assumption behind your plan is not our view of our own results. |
| Inflation | 3% a year, 1% volatility | Firm assumption. |
| Federal income tax | 2026 brackets and standard deduction | Rev. Proc. 2025-32, read from the IRS release itself. The brackets rise with the price level in the projection, because they do in law. |
| Tax on Social Security | Up to 85% taxable | IRC §86. These thresholds are statutory and have never been indexed, so more of every cohort's benefit becomes taxable over time. The projection does not inflate them. |
| Drag on the taxable account | 0.8% a year | Firm assumption, standing in for tax on dividends and turnover. |
After three answers the tool shows how long what you have covers what you want to spend, on its own — no growth, no Social Security, no pension, nothing further saved. It shows that instead of a percentage on purpose. Before it knows about Social Security and what you save each year, a percentage would describe a household with neither, and that is a different household rather than a rough version of yours. On our own test figures the gap between those two answers was 7.9% and 94.9%.
The tool stops rather than estimating when it meets something it does not model: a second property and its mortgage, self-employment income and its payroll tax, a lump sum of banked time, or forced minimum distributions. It says which one and produces no figure. A confident answer on a basis you cannot see is indistinguishable from a right one.
It does not model required minimum distributions. If you have a large pre-tax balance, the tax figure it shows is understated, and the effect grows the longer the plan runs. It also does not model state income tax on your earnings before retirement, health costs that rise faster than general inflation, long-term care, or a survivor's benefit election.
It does not check that what you say you save is affordable out of what you earn. If you tell it you earn $60,000 and save $50,000, it will model exactly that, indefinitely. It flags the mismatch on screen, and then it does what you asked.
The arithmetic runs on your own device. There is no account, no server, and no database. We do not receive what you type and could not produce it if asked. Your answers are stored in the web address of the page itself, which is why bookmarking it brings your figures back — and why anyone you send that link to can see them.
The same household model exists twice: once in Python, which is the source of truth, and once in the JavaScript that runs on your device. Every deploy checks them against each other — the tax tables, the projection itself on a shared random-number generator, the refusals above, and every sentence the page writes back to you, character for character. At the last run that was 237 separate checks. If any one disagrees, the deploy stops.
The general version builds your full Retirement Plan. If you are covered by a New York City pension, the department versions start from your own plan rules rather than asking you to look them up: NYPD, FDNY. The pension side of those is documented separately in how the Retirement Snapshot works.
The builder is an educational tool. It does not recommend a course of action, a product, or an investment, and no adviser reviews what it produces. Sirmium Capital LLC is a New York state-registered investment adviser and is not registered with the SEC; nothing here is an offer or solicitation in any jurisdiction where we are not registered or exempt. A financial plan prepared by Sirmium Capital is provided only under a signed advisory agreement. See our full disclosures.