Risk & Market Behavior

One Average Says You're Fine. A Thousand Futures Disagree.

What a Monte Carlo retirement simulation actually does, why the tidy average-return projection most calculators show you can be dangerously optimistic, and how to read a success rate without fooling yourself.

11 min readLast reviewed July 2026
The short version
  • Most free retirement calculators assume your portfolio earns the same average return every year. Real markets don't do that — and when you're withdrawing, the order of good and bad years matters as much as the average.
  • A Monte Carlo simulation runs your plan through 1,000 different possible sequences of market returns and reports how many futures your money survives. That's your success rate.
  • The gap is not academic. In the worked example below — a real run of this site's engine — the steady-average projection ends at age 92 with $1.2 million to spare. The simulation says 371 of 1,000 futures run out of money. Same plan, same assumptions.

The retirement math most calculators do

Take your savings, add your contributions, grow the pile by 6 or 7% a year, subtract your spending in retirement, and see if the line stays above zero until some age. That's the math inside most free retirement calculators, and it isn't wrong, exactly — it's the right way to get a first rough answer, and it's fast.

But it quietly assumes something no market has ever done: that returns arrive in the same smooth amount, every year, forever. No 2008. No lost decade. No brutal first three years of retirement. Just a tidy compounding curve. The projection it produces looks precise — a single dollar figure at every age — and that precision is exactly what makes it misleading. It's one future out of thousands the same plan could actually live through, and it happens to be one of the gentler ones.

The same plan, run both ways

Here's a concrete case, computed with the same engine that runs this site's app — not estimated, not rounded from someone else's blog. Meet a 55-year-old with a solid, ordinary plan:

  • $720,000 saved today — $500k in a 401(k), $60k Roth, $120k brokerage, $40k cash savings
  • Saving $15,000/year until retiring at 65
  • Social Security of about $2,600/month (today's dollars), claimed at 67
  • Spending goal of $6,000/month in retirement, planning to age 92
  • 7% average return before retirement, 5% after, 3% inflation

Run the steady-average projection and the news is great: the portfolio grows to about $1.62 million by 65, funds every year of spending, and still ends at age 92 with roughly $1.2 million left over. Plan works. Close the tab, book the trip.

Now run the exact same plan through 1,000 simulated market sequences — same average returns, same inflation, same spending, with the only change being that returns arrive in realistic random orders instead of one smooth line:

How it's runResult at age 92Read
Steady average, one path$1.2M remainingLooks comfortably safe
1,000 simulated sequences629 survive · 371 run out62.9% success — roughly 1 in 3 futures fails

Same plan. Same average assumptions. The steady-average view says "you'll die with a million dollars." The honest view says the odds are closer to 2 in 3 — and in the futures that fail, the money typically runs out around age 87, with the earliest failures hitting at 76.

An average-return projection isn't a forecast of your retirement. It's one gentle future wearing the costume of certainty.

Why the order of returns matters more than the average

The mechanism behind that gap has a name — sequence-of-returns risk — and one honest sentence explains it: when you're withdrawing money, a bad market early does far more damage than the same bad market late.

If a 30% crash lands in year two of retirement, you're forced to sell into the hole to pay for groceries, and the shrunken portfolio then has 25 years of withdrawals still ahead of it. If the identical crash lands in year twenty, it hits a portfolio that already had two decades of growth behind it and far fewer years left to fund. Same crash, same average return over the whole retirement — completely different outcome.

You can see how wide the range really is in the simulation above. Among those 1,000 futures — every one using the same average-return assumptions — the 90th-percentile outcome ends age 92 with about $3.9 million, the median with about $527,000, and the 25th percentile at zero. Nothing differs between those futures except the order the years showed up in. That spread is sequence risk, made visible.

This is also why two honest tools can disagree about the same plan — a difference in how each one models variance produces different success rates from identical inputs. We wrote up a side-by-side of how the major calculators differ in Why Retirement Calculators Disagree.

What a success rate actually means — and doesn't

A Monte Carlo success rate is the share of simulated futures in which your money outlived your plan. 62.9% means 629 of 1,000 futures made it. It is a confidence level, not a grade — and definitely not a guarantee. A few honest reading rules:

  • 90%+ is generally strong. One future in ten still struggled — but for most plans with any spending flexibility, this is solid footing.
  • The 80s are workable with flexibility. If you could genuinely cut spending 10-15% during a bad stretch, an 85% plan is a very different thing than an 85% plan with zero slack.
  • The 70s deserve attention. Not panic — attention. Usually one lever (retirement age, spending, Social Security timing) moves this meaningfully.
  • Below 70% is leaning on luck. The plan works only if the market is kind early. That's a hope, not a plan.
  • 100% doesn't exist. Any tool showing 100% has simply stopped looking for failure modes. There's always a future bad enough — the question is whether it's plausible enough to plan around.

One more honesty rule that gets skipped a lot: a success rate describes the futures the simulation modeled. It can't price in a tax law that doesn't exist yet, a health event with no precedent, or a market regime unlike anything in the data it draws from. Projection, not prediction — every number on this site carries that caveat, on purpose.

Try it — how fragile is your plan to a bad early sequence?

What moves the number

The uncomfortable and empowering thing about a success rate is how sensitive it is to the levers you actually control. Here's the same 55-year-old's plan at four different spending levels — nothing else changed, every run from the same engine:

Same plan, four spending levels — hover a bar
Monte Carlo success rate falls as monthly spending rises The same retirement plan run at four monthly spending levels: $5,200 a month succeeds in 88.4% of 1,000 simulated futures. $5,600 succeeds in 76.0%. $6,000 succeeds in 62.9%. $6,400 succeeds in 48.8% — a coin flip. An $800 monthly difference separates a strong plan from a coin flip. 0% 20% 40% 60% 80% 100% 90% — generally considered strong footing $5,200/mo $5,600/mo $6,000/mo $6,400/mo 88.4% 76.0% 62.9% 48.8%

Every bar is the same $720k-at-55 plan from the worked example — only the monthly spending goal changes. $1,200 a month of spending separates strong footing (88.4%) from a literal coin flip (48.8%). That sensitivity cuts both ways: it's why an over-optimistic calculator is genuinely dangerous, and why modest, early adjustments move the odds so much.

Spending is only one lever. Retiring a year or two later, claiming Social Security on a smarter timeline, or shifting when the portfolio takes its withdrawals each move the success rate too — usually by more than people expect, and sometimes in directions they don't. Finding which lever moves your plan most is precisely what simulation is for.

Monte Carlo vs. historical back-testing

Monte Carlo isn't the only honest way to test a retirement plan, and it's worth being straight about the tradeoff. A Monte Carlo simulation generates futures statistically — which means it explores sequences history hasn't produced yet, but leans on assumptions about how returns behave. Historical back-testing does the opposite: it replays your plan against every actual market start-year on record — retiring into 1929, into 1973's stagflation, into 2008 — no statistical assumptions, but limited to sequences that have already happened once.

They're complementary, and a rigorous plan checks both: the simulation for breadth of futures, the historical record for grounding against the real catastrophes. This site's engine runs both — the Monte Carlo simulation described here and a historical back-test against market data back to 1928 — because a plan that passes one and fails the other is telling you something important.

How this engine runs it

Numbers in this article aren't illustrations — they're output. The specifics, so you can judge the tool rather than take our word:

  • 1,000 simulated futures per run, with randomized return sequences built from historical market behavior, inflation applied throughout.
  • The full plan is modeled inside every future — federal and state taxes, Social Security timing (including spousal and survivor mechanics), required minimum distributions, and tax-aware withdrawal ordering across account types. A simulation that skips taxes and RMDs isn't simulating your plan; it's simulating a simpler one.
  • Reproducible by construction. The simulation is seeded — the same plan produces the same success rate every time, which is what makes the engine testable at all.
  • Verified before every release: 27 benchmark scenarios re-run at zero-drift tolerance through a 19-gate pipeline. The whole regime is public in the trust framework, and the complete math is documented in the methodology.

Common questions

What is a Monte Carlo retirement simulation?
Instead of assuming your portfolio earns the same average return every single year, a Monte Carlo simulation runs your retirement plan through many randomized sequences of market returns — 1,000 in this engine — and reports how many of those futures your money survives. The output is a success rate: the share of simulated futures in which the plan held up. It captures the one thing an average hides: the order returns arrive in matters enormously when you're withdrawing.
What Monte Carlo success rate should I aim for?
There's no universal number, but a useful frame: 90%+ is generally considered strong; the 80s are workable if you have real flexibility to cut spending in a bad market; the 70s deserve attention; below 70% the plan is leaning heavily on luck. The right target depends on how flexible your spending is and how much of your income floor is guaranteed (Social Security, pensions). A success rate is a confidence level, not a guarantee — 95% still means 1 future in 20 ran out.
How many simulations are enough?
Somewhere around 1,000 runs, the success rate stabilizes to within about a percentage point — tighter than the precision of any assumption you're feeding it. More runs sharpen the number slightly but change no decisions. What matters more than the count is what each run models (taxes, Social Security, RMDs), and whether the simulation is reproducible rather than giving a different answer every refresh.
Is Monte Carlo better than historical back-testing?
They answer different questions. Monte Carlo asks: across a wide range of statistically plausible futures, how often does the plan survive? Back-testing asks: would the plan have survived every actual market sequence on record — including 1929, the 1970s, and 2008? Monte Carlo explores futures history hasn't produced yet; back-testing keeps the simulation honest against sequences that really happened. A rigorous plan checks both, which is why this engine runs both.

The worked example above is one plan. The full app runs yours — 1,000 futures, Social Security timing, taxes, RMDs, and a stress test against real market history. Free, no signup.

Run your plan through 1,000 futures →