How this works, and where the numbers come from
Coastfarer simulates 100,000 versions of a financial life and reports what tends to happen. "On track" means one thing throughout: the money lasts through the plan horizon (age 90 by default). Everything is in today's dollars, with inflation already taken out.
Coast FIRE
Coast FIRE is the point where what you've already invested can grow into your retirement on its own, without another penny added.
After that point you still need to cover today's bills, but you no longer need to save for retirement, so a lower-paid or lower-stress job becomes an option. The site is named for it, and answers the question directly: under the levers, "coast from here" re-runs every life with one change, that from today you earn just what you spend (plus any debt payments) and add nothing further to your investments. Retirement spending stays exactly as you set it, pay stops growing, and Social Security is re-estimated from those lower earnings.
Reaching it isn't a single number here, it's a share of outcomes: what you see is how many of the simulated lives never run out if you stop saving today, next to how many do if you carry on. The gap between those two is how much your future still depends on what you add from here.
The example you land on
A 35-year-old on roughly the national median income, saving steadily. Not a target, and not anyone in particular.
Before you enter anything, the page runs a worked example: 35 years old, $60,000 saved, $80,000 gross, spending $48,000, retiring at 65. Those numbers are anchored rather than invented. Median US household income was $83,730 in 2024 (Census, P60-286), and the median retirement-account balance for households aged 35 to 44 who have one was $45,000 in the 2022 Survey of Consumer Finances. The example sits a little above that balance, because somebody looking up a tool like this is usually already saving.
It lands near a coin flip on purpose. An example that always succeeds would make the model look better than it is and make your own numbers feel worse by comparison; one that always fails would say nothing about what changes anything. Replace it with your own figures and none of it carries over.
The market model
We replay real market history instead of assuming a smooth average return.
Each simulated life draws its investment returns by replaying blocks of real market history: inflation-adjusted US large-cap annual returns from 1928 to 2025 (Damodaran's S&P 500 total-return series, deflated by BLS CPI). They're resampled in multi-year blocks, so crashes stay streaky and recoveries stay slow, the way they actually were. History's average is then shifted down to a conservative 5% real forward assumption. Keeping the fat tails and the bad decades is why outcomes spread as widely as they do; a smooth "average return" hides all of it.
Taxes
Real 2025 tax law, and we skip the deductions that would flatter you.
In gross mode, each year of each life is taxed with 2025 law: federal single-filer brackets, FICA (Social Security to the $176,100 wage base + Medicare incl. the 0.9% additional), and a representative top-marginal state rate. No deductions or credits are modeled. Skipping the standard deduction makes the tax bite slightly worse than reality, which is the safer way to be wrong.
Social Security
We assume it pays about three-quarters of what's scheduled, rather than all of it or none of it.
Estimated with the SSA's actual formula: earnings capped at the wage base, averaged over a 35-year career, run through the 2025 bend points, adjusted for claim age (62 to 70). The default then pays 75% of the scheduled benefit. That figure comes from the SSA's own Trustees, who project the trust fund depletes around 2033, after which ongoing payroll taxes still cover roughly 75% to 79% of benefits. Assuming it pays nothing at all is too pessimistic; about three-quarters is the honest do-nothing case. You can set full, reduced, or none.
Life events, calibrated to real data
Job losses, medical bills and the rest happen at rates taken from public data. Job loss is likelier in a crash year, which is what makes it bite.
The events that happen to you carry base rates and magnitudes from primary US sources rather than ballpark guesses. Job loss also couples to the market: it's three times likelier in a crash year, so simulated people lose jobs while their portfolios are already down, which is how real recessions work. The events you choose (marriage, kids, home, sabbatical, and so on) use your own answers from the "Life plans" panel.
| Event | Rate & effect | Primary source |
|---|---|---|
| Job loss | 2.5%/yr, ×3 in a crash; a year at $0 then a ~5% wage scar | BLS Displaced Workers, JOLTS |
| Severe pay cut | 0.4%/yr, ×2 in a crash; −15% for 3 years | NBER WP 25470 |
| Medical shock | 3→8%/yr rising with age; −$12k | ACA OOP max, AHRQ MEPS |
| Disability | 0.3→1.3%/yr rising with age; −50% income | SSA actuarial |
| Inheritance | 2%/yr in the receiving window; +$120k (conservative) | Penn Wharton, Fed FEDS note |
| Divorce | 1.8%/married-yr, front-loaded; −$30k + household reversion | CDC/NCHS, IFS |
| Eldercare | 4.5%/yr mid-life; −$8.5k/yr for 5 years | AARP |
| Lawsuit | 0.2%/yr; −$40k. Deliberately unsourced | no credible base rate exists; kept small |
How to read the sourcing: some of these are direct statistics, others are reasoned conversions (a lifetime share turned into an annual rate). The lawsuit rate has no defensible source at all: the widely-quoted "10% a year" traces to insurance marketing, so we keep it small and say so. Every figure the model uses is in the table above, with its source; there is no second set of numbers held back.
Known limits, stated plainly
- Single-filer taxes only (married-filing-jointly is a known gap; single is the conservative default).
- Withdrawals aren't taxed separately (no LTCG vs. ordinary distinction yet).
- Divorce models legal costs, not the asset split.
- Event magnitudes are point estimates, not distributions.
- Age ramps are linear where real incidence curves accelerate.
- One market model (US large-cap history); no asset-allocation choices yet.
- Career pay follows one median curve (three settings), so it can't represent your occupation or field. Real earnings growth is also skewed toward bad surprises.
- A mortgage's payment is modeled; the house is not. No equity and no sale, so a homeowner's total wealth is understated here.
Reproducibility
The simulation is deterministic: the same inputs and the same seed produce the same 100,000 lives, to the dollar. That's why the numbers on the page don't wobble between visits, and why a shared link reproduces someone else's results exactly.
That holds within one version of the model. When the model changes (a recalibrated rate, a corrected definition), the same inputs can produce different numbers. Every link records the version it was made with, so a link from an older version still loads its inputs and tells you the numbers may differ from what the sender saw.
Privacy
Every simulation runs in your browser, and your numbers are never sent anywhere. The site's Content-Security-Policy (connect-src 'self') forbids the page from talking to any other server. Your browser enforces that, so it isn't something you have to take on trust. No ads, no accounts, no cookies.
What's counted
To tell whether real people use this (most of a new site's traffic is bots), the page sends one word to this site when you do certain things, at most once per visit: that the page loaded, that you ran a scenario, opened the input panel, copied or arrived by a share link, came back to a saved scenario, traced a life, sliced the lives, or clicked feedback. That word is all that's sent.
What's kept is a daily total per event, like "21 September: 14 visits ran a scenario", and nothing else. No IP address, no browser details, no identifier, no cookie, no time finer than the day, and never your numbers. There's no record of any single visit to look up, and nothing to link one visit to another.
It's off if your browser sends Global Privacy Control or Do Not Track. You can also switch it off here, for this browser:
The other exception is feedback, and only if you choose to send it. Send feedback opens a form on Tally, a separate service, in a new tab. Nothing from this page goes with it: not your numbers, and not even which page you came from. It's anonymous unless you leave contact details yourself.
This is not financial advice. Coastfarer is an educational simulation. It models one simplified version of the world with the assumptions above, and reality will differ. Nothing here accounts for your full situation. For decisions, consult a licensed professional.