S&P 500 6,910.39 ↓ 3.7%Dow Jones 48,147.33 ↓ 1.93%Nasdaq 22,617.73 ↓ 5.28%BTC $60,741 ↑ 3.4%ETH $1,705 ↑ 2.7%EUR/USD 1.099Inflation 3.5% YoYLive market dataS&P 500 6,910.39 ↓ 3.7%Dow Jones 48,147.33 ↓ 1.93%Nasdaq 22,617.73 ↓ 5.28%BTC $60,741 ↑ 3.4%ETH $1,705 ↑ 2.7%EUR/USD 1.099Inflation 3.5% YoYLive market data

Monte Carlo Retirement Simulations: What That Success Score Really Means

The probability of success number on your retirement plan is not a crystal ball. Here is what it measures, what it quietly leaves out, and how to read it without scaring yourself or fooling yourself.
Monte Carlo Retirement Simulations: What That Success Score Really Means

Key takeaways

  • A Monte Carlo simulation runs your retirement plan through thousands of randomly ordered return sequences, so it can test the one thing a single average return cannot: the risk of hitting bad markets early while you are withdrawing.
  • The headline 'probability of success' is just the share of simulated paths that met the model's success test under its assumptions. An 87 percent score means 87 of 100 simulated paths passed, not a measured probability of what will happen to you.
  • In the common fixed-spending version of the model, a 'failure' assumes you never adjust. Many tools can instead model dynamic spending and guardrails, so the fixed-spending failure rate is not the whole story.
  • Research by Derek Tharp on Kitces.com shows that if you recalculate spending every month to hold a constant probability of success, downward adjustments get triggered in nearly every historical scenario. But Tharp cautions that steering to a constant probability is itself unwise, so that figure reflects the method, not what a careful household must endure.
  • A 100 percent score means every sampled path passed under your assumptions. It cannot establish zero risk, prove you are being frugal, or guarantee a large estate, because the model samples only the futures its assumptions allow.
  • Tail risk cuts both ways and depends on frequency: daily and monthly returns have fatter tails than a bell curve, yet one Kitces.com study of a 60/40 portfolio found annual returns looked roughly normal, while long-horizon retirement outcomes can be overstated when mean reversion is ignored.
Send this guide

It goes to your followers on the network you pick.

If you have ever sat down with a retirement calculator or an advisor's software, you may have walked away with a single number stuck in your head. Something like, "Your plan has an 87 percent chance of success." It sounds precise. It sounds reassuring, or maybe a little alarming, depending on the number. And it leaves most people with the same quiet question: a chance of what, exactly?

That number almost always comes from a Monte Carlo simulation. The name is borrowed from the famous casino town, because the method leans on randomness the way a roulette wheel does. It is one of the most common tools in retirement planning, and also one of the most widely misread. People treat the score like a weather forecast or a report card grade. It is neither. Understood properly, it is a useful way to stress test a plan against an uncertain future. Understood poorly, it can push you into needless worry or false comfort.

Let us walk through what a Monte Carlo simulation actually does, what the success percentage really means, and how to read your own number without reading into it things that are not there.

The problem it was built to solve

Start with the simpler tool that came first. A basic retirement calculator asks for a handful of inputs and assumes a single, smooth rate of return. You tell it you expect 6 percent a year, and it grows your money by 6 percent every single year, like clockwork, then subtracts your withdrawals and tells you whether the money lasts.

The trouble is that markets have never once behaved that way. Real returns arrive in a jagged, unpredictable order. One year is up 20 percent, the next is down 15, then flat, then up 8. The long run average might genuinely be 6 percent, but the path there is bumpy, and the bumps matter enormously once you are pulling money out instead of putting it in.

This is the heart of what planners call sequence of returns risk. Two retirees can earn the exact same average return over 30 years and end up in completely different places, simply because one of them hit a bad stretch early while taking withdrawals and the other hit it late. A straight line calculator cannot see this risk at all, because it never lets returns vary. Monte Carlo was built specifically to put that variation back in.

How the simulation actually runs

The mechanics are less mysterious than the name suggests. Instead of running your plan through one assumed future, the software runs it through thousands of different possible futures, then counts how many of them work out.

You feed it the same raw inputs you would give any planner: your current savings, how much you are adding each year, when you plan to retire, how much you expect to spend, and how long the plan should last. The important addition is that you also give it an expected average return and a measure of how much returns tend to bounce around from year to year, which statisticians call volatility.

From there the program does something a person never could by hand. For a single simulated lifetime, it generates a random return for year one, pulled from a realistic range. Then a random return for year two. Then year three, all the way to the end of your plan. That gives you one complete, plausible path through retirement, with its own particular sequence of good and bad years. Then it throws that path away and does the whole thing again with a fresh set of random years. And again, and again. A typical run tests many thousands of these paths. Industry explainers often describe a thousand paths as a practical minimum and ten thousand as common, because you need enough paths for the percentages to settle into something stable.

At the end of each simulated path, the program checks one thing: did the portfolio meet the model's success test, which in most tools means the money lasted to the planned end age without hitting zero? A path that finishes with money left over counts as a success. A path that ran out counts as a failure. Tally up the successes, divide by the total number of paths, and you get the headline figure. An 87 percent success rate simply means that in 87 out of every 100 simulated futures, under that model's assumptions, the portfolio met the test.

What the success percentage does and does not mean

Here is the single most important thing to understand, and the part the software rarely spells out clearly. A success rate is a count of simulated paths that met one specific test inside one specific model. It is not a measured probability of what will happen to you, and it is not a verdict on whether your plan is affordable.

A lot of the confusion comes from how the common version of the model handles spending. The default setting in many tools assumes you keep spending exactly the same inflation adjusted amount every year, no matter what the market does, right down to a zero balance. That is a modeling choice, not a law of the method. It is worth being precise here, because a common myth says Monte Carlo cannot account for the fact that real people adjust. That is not true. Plenty of modern tools can simulate dynamic spending, floors and ceilings, and formal spending guardrails that raise or trim withdrawals as the portfolio moves. The point is narrower: in the fixed spending version of the model, a "failure" assumes a retiree who never once changed course.

So when a fixed spending plan shows a 13 percent failure rate, that is not a 13 percent chance you end up destitute. In most of those failing paths, what actually happened is that a bad market stretch would have called for some kind of adjustment along the way. The financial planner Michael Kitces has argued for years that "probability of success" is a misleading label, and that these outcomes would be more honestly framed as a probability of adjustment.

But that relabeling comes with a caveat that often gets dropped, and it is easy to twist in either direction. In a 2022 analysis on Kitces.com, lead researcher Derek Tharp showed that calling the failure rate a "probability of adjustment" can understate how often adjustments get triggered, but only under one very particular method. Tharp stepped a hypothetical retiree through market history going back to 1871, recalculating spending every single month to hold the probability of success at a constant level, and he counted any downward change below the starting inflation adjusted amount, even a trivial one, as an adjustment. Measured that way, a plan pinned at a constant 95 percent still logged a downward adjustment in roughly 96 percent of the historical scenarios, and one pinned at 70 percent in almost all of them. Here is the part that often gets lost when that statistic travels: Tharp does not present this as how a careful household should plan. He explicitly cautions that steering to a constant probability of success is probably not wise in the first place, precisely because it forces constant, often tiny tweaks. So the finding is narrower than it sounds. It says that if you chase a fixed score by re-solving every month, you will fiddle with spending almost continuously, which is an argument for guardrails that build in a buffer, not proof that an ordinary 95 percent plan hides near-certain hardship.

The deeper point still holds, though. No one lives with hindsight. You cannot know in the moment whether you are in a path that was going to recover on its own, so a cautious person may trim spending preventively. A failing path also does not prove that a few modest, painless cuts would have been enough. The honest questions are when the cuts would be needed, how deep they would have to go, and whether they are actually feasible for your household. A shortfall that shows up at 85 with large required cuts is a very different problem from one that shows up at 72 needing a small, temporary trim.

The flip side is worth holding too. The model counts a path as a success whether you finish with one dollar or with several million, and it makes no distinction between them. That same Kitces.com research points out that at high scores, many of those "successful" paths leave a large unspent balance, which is why it also frames a high success rate as a kind of probability of excess. That is a real tradeoff, just a gentler one than running short. But notice what it is and is not: it is a tendency across many high scoring paths, not a promise about your particular plan.

How to read your own number

Because the score is a model dependent count rather than a literal survival odds, chasing 100 percent is usually the wrong instinct. It helps to be careful about what a 100 percent result actually tells you. It means every path the model sampled met its success test under the assumptions you fed it. That is genuinely informative, but it is also limited. It cannot establish that your plan carries zero risk, because the model can only sample from the futures its own assumptions allow, and the real world can serve up conditions outside that range. It cannot prove you are being unnecessarily frugal, and it does not guarantee a large estate at the end. All of those depend on the return, inflation, and spending assumptions baked in. A 100 percent score is best read as an invitation to look harder at the inputs and at whether a plan this conservative matches how you actually want to live, not as proof of either safety or waste.

Where should a reasonable target land? There is no single right answer, and this is a place where honest sources disagree. Different providers set different comfort ranges, and a fair amount of planner commentary gravitates toward a band in the 80s, with some accepting lower and some wanting higher depending on the household. What matters more than the exact cutoff is that a percentage by itself cannot tell you whether a plan is affordable or how severe a shortfall would be. Someone whose essential bills are fully covered by Social Security and a pension can often tolerate a lower score, because the exposed spending is discretionary. Someone leaning almost entirely on a portfolio generally wants more cushion. The number is one input to that judgment, not the judgment itself.

Read the band your number falls in as a prompt, not a grade, and treat any tidy interpretation as illustrative and tied to the particular model that produced it. A plan sitting at 72 percent is not doomed, but it is telling you to look closely at the levers you control: spending a bit less, working a year or two longer, or delaying Social Security. On that last point, be precise about what delaying does. Waiting past full retirement age increases your monthly benefit through delayed retirement credits, and those credits keep accruing only up to age 70, after which there is no further increase for waiting. A larger monthly check is not the same as a larger lifetime total, though. Whether delaying raises what you collect over your whole life depends on how long you live, so it is a tradeoff to weigh against your own health and income needs, not a guaranteed win.

One more thing worth knowing. Not all Monte Carlo engines are built the same, and different software can hand you different scores for the exact same plan. The results depend heavily on the assumptions baked into the model: what average return it uses, how much volatility it assumes, whether it treats inflation as a random variable or a fixed number, whether spending is fixed or dynamic, and how it handles taxes and fees. Two reputable tools can disagree for the same plan, sometimes by a wide margin. The number is a property of the plan and the model together, never the plan alone.

What actually moves the number

Once you accept that the score depends on assumptions, you start to see how sensitive it can be. Small changes to the inputs can swing the result in ways that feel out of proportion, which is exactly why it pays to understand the drivers rather than fixate on the headline.

The biggest lever is usually the return assumption, and it is also the one people fiddle with most casually. Nudge the expected annual return down by a single percentage point, and a plan that looked comfortable can suddenly look shaky. The chart below shows a purely illustrative example of how one hypothetical plan's success score might slide as the assumed return changes. The numbers are invented to show the shape of the relationship, not to predict any real plan.

Spending is the second great lever, and often the one you have the most direct control over. Because withdrawals compound against you the same way returns compound for you, a modest cut in planned spending can lift a success rate noticeably, especially in the early retirement years when sequence risk bites hardest. This is why so much modern retirement thinking has moved toward flexible withdrawal approaches, such as the guardrails mentioned earlier, that raise or trim withdrawals as the portfolio moves. When a tool is set up to model those adjustments on purpose, rather than assuming rigid spending, the exercise shifts from "will this survive untouched" to "what would I change, and when." That is a more useful question, as long as you remember that the adjustments have to be ones you could actually live with.

Time horizon is the third lever. The longer the plan has to run, the more years of randomness it has to survive, and the lower a given plan's score tends to be. Plan to age 90 and your number looks one way. Plan to 100, a reasonable hedge against the real possibility of a long life, and the same plan will score lower simply because it is being asked to do more. A conservative horizon is prudent, but it is worth remembering that a scary number partly reflects how cautious you told the model to be.

Where the model falls short

Monte Carlo is a genuine improvement over a single straight line, but it is still a model, and every model simplifies. A few of its blind spots are worth keeping in front of you.

First, it is only as honest as its inputs. Feed it an optimistic return assumption, understate your spending, or forget about taxes and healthcare, and it will cheerfully hand you a reassuring score built on sand. A percentage reported to the decimal point can lull you into trusting inputs that were really just educated guesses. The old rule holds: a tidy output cannot rescue messy assumptions.

Second, the way the engine generates random returns is itself an assumption about how markets behave, and this is a place where a popular criticism needs care. Many engines draw from a smooth bell curve. At short frequencies, a single day or month, real markets genuinely do have fatter tails than that curve suggests, so sharp crashes and spikes turn up more often than the tidy math predicts. But most retirement projections run on annual returns, and at that scale the picture changes. Testing a 60/40 portfolio on Robert Shiller's data going back to 1871, Derek Tharp's work on Kitces.com found that the annual returns did not show a statistically significant departure from a normal distribution, so the fat tails that appear at the daily and monthly level largely fade at the yearly frequency he studied. It is tempting to conclude that retirement simulations always understate long-term downside anyway, but his research does not support that leap. Over multi-decade horizons the opposite can happen. Because standard Monte Carlo treats each year as independent and ignores the way markets have historically tended to mean revert, with bad stretches often followed by recoveries, it can string together long runs that are worse than anything in the historical record. In that analysis, which modeled a 60/40 portfolio taking inflation adjusted withdrawals and compared ten thousand simulated paths against overlapping 30 year periods in U.S. history from 1871 to 2015, about six and a half percent of the simulated 30 year paths finished worse than the worst 30 year stretch in that historical record. So whether a model overstates or understates tail risk depends on the return frequency and the model's assumptions. Daily and monthly tails are understated by a bell curve, annual returns in that study looked roughly normal, and long-horizon retirement outcomes can be overstated when mean reversion is left out. Either way, a 100 percent result still cannot be read as zero risk, because the real future can differ from whatever distribution the model used.

Third, the handling of spending is a modeling choice you should check rather than assume. If the tool uses fixed inflation adjusted spending, the plain success rate will look more pessimistic than a flexible retiree's real odds. If it uses dynamic spending or guardrails, the headline may look friendlier, but then the relevant question becomes how large and how frequent the modeled cuts are, since a high score can hide a lot of belt tightening.

Fourth, the score collapses a rich, messy future into one number, and a single number cannot tell you how deep the failures went or how large the leftover balances were. Two plans can share the same 85 percent score while being wildly different underneath, one with shallow, easily fixed shortfalls and one with a handful of catastrophic ones. The headline hides all of that.

Putting it to work

None of this means you should dismiss the number. It means you should use it the way a pilot uses an instrument, as one reading among several, valuable precisely because it tests things your gut cannot. Here is a sensible way to hold it.

Treat the score as a conversation starter, not a conclusion. If it comes back low, resist the urge to panic or to crank up your return assumption until the number looks friendly. Instead, pull the real levers one at a time and watch how each moves the result: a later retirement date, a slightly trimmer budget, a Social Security claiming decision, a more realistic spending plan in the early years. Seeing which lever moves the number most tells you where your plan is actually fragile. And when a plan leans on future adjustments, press on whether those cuts are ones you could realistically make, how big they would be, and when they would land.

Run it again when life changes, not just once. A Monte Carlo result is a snapshot of a single moment's assumptions. A market swing, a move, a health event, a shift in your spending, or simply getting a few years closer to the finish line can all change the picture. The number is most useful as something you revisit, watching the trend across years rather than clutching any single reading.

And pair it with the simpler mental checks that got people through retirement long before this software existed. The rough arithmetic behind the 4 percent rule, a clear sense of which bills are covered by guaranteed income, and an honest look at how willing you really are to cut back in a bad year. The simulation adds rigor to those instincts. It does not replace them.

The small tool below lets you play with the plain compounding math behind a growing nest egg. It is not a Monte Carlo engine, so it assumes a single smooth return the way the old calculators do. That is the point. Compare the clean line it draws to the jagged reality we have been describing, and you will feel, in your gut, exactly why planners reach for the messier, randomized version when the stakes are your actual retirement.

In the end, the success score is a humble thing dressed up in precise clothing. It cannot tell you what will happen, and it cannot certify that a plan is safe or affordable on its own. What it can do is show you how much weather your plan appears to handle under a specific set of assumptions, and nudge you to decide, calmly and in advance, how you would steer if the skies turned rough. A plan you have stress tested and know how to adjust will almost always serve you better than a plan that merely looked certain on paper. The number is not the answer. It is a better question.

Your earning years are the engine

Retirement math is career math in disguise.

Contribution rates matter, but the salary they multiply against matters more. Whether you are mid-career or planning a second act, RealWorldCareers shows which work fits your brain so your strongest earning years are actually your strongest.

Find the career your brain was built for
RealWorldCareers is built by our parent company, Advanced Learning Academy. Same family, same standards.

Questions people ask

Is a higher Monte Carlo success rate always better?

Not necessarily. A very high score, especially 100 percent, means every path the model sampled passed its success test under the assumptions you entered. That is useful information, but it cannot prove the plan is risk free, because the model only samples the futures its own assumptions allow, and it does not prove you are spending too little or that a large balance will be left over. Target ranges also differ by provider, so treat any single cutoff as model dependent. What matters more is whether the inputs are realistic and how much of your spending is already covered by guaranteed income like Social Security or a pension.

Does a failing simulation mean I will go broke?

No, and this is the most common misunderstanding. In the common fixed-spending version of the model, a failure assumes you keep spending the same inflation adjusted amount every year no matter what, down to a zero balance. Real retirees adjust, and many tools can model those adjustments too. So a fixed-spending failure usually represents a future where you would have needed to change course at some point, not one where you literally end up destitute. Be careful with that reframing, though. Research by Derek Tharp on Kitces.com shows that if you recalculate spending to hold a constant probability of success, adjustments get triggered in nearly every scenario, but Tharp himself warns that planning to a constant probability is not wise. The practical questions are when a real adjustment would hit, how deep it would be, and whether it is feasible for you.

Can Monte Carlo tools account for the fact that people cut back in bad markets?

Yes. It is a myth that Monte Carlo must assume rigid, constant spending. That is only the default setting in some tools. Many modern engines can simulate dynamic spending, spending floors and ceilings, and formal guardrails that raise or trim withdrawals as the portfolio moves. The catch is that a friendlier-looking score from a dynamic model can hide frequent or sizable spending cuts, so it is worth asking how large and how often the modeled adjustments are, not just what the headline percentage says.

Why do two tools give me different success rates for the same plan?

Because the score depends on the assumptions inside the engine, not just on your plan. Different tools use different expected returns, different assumptions about volatility, different ways of modeling inflation, different spending rules, and different handling of taxes and fees. Those choices can move the result meaningfully for the same plan. The number is a property of your plan and the model together, so it is most useful for comparing scenarios within one consistent tool rather than across different ones.

Does a bell-curve Monte Carlo understate the risk of a crash?

It depends on the time frame. At short frequencies, a single day or month, real markets do have fatter tails than a smooth bell curve, so extreme moves happen more often than the simple math suggests. But most retirement projections run on annual returns, and that is a different story. Testing a 60/40 portfolio on data back to 1871, Derek Tharp's research on Kitces.com found that the annual returns did not depart significantly from a normal distribution, so those short-frequency fat tails largely fade at the yearly scale. And it does not follow that long-horizon simulations always understate downside either. Because standard Monte Carlo treats each year as independent and ignores the tendency of markets to mean revert, it can actually generate multi-decade outcomes worse than anything in the historical record. In that study, which modeled a 60/40 portfolio taking inflation adjusted withdrawals against overlapping 30 year periods in U.S. history from 1871 to 2015, about six and a half percent of simulated 30 year paths were worse than the worst 30 year stretch in that record. So the direction of the bias depends on the return frequency and the model's assumptions.

Does delaying Social Security guarantee I collect more overall?

No. Delaying past your full retirement age increases your monthly benefit through delayed retirement credits, and those credits only accrue up to age 70, after which waiting adds nothing. A bigger monthly check is not the same as a bigger lifetime total. Whether you collect more across your whole life depends on how long you live, so it is a tradeoff to weigh against your health, other income, and cash flow needs rather than an automatic win.

Just so you know: DollarFlourish is an educational publisher, not a financial, tax, or investment advisor. Numbers and rates change. Verify anything important with a licensed professional before acting on it. Some links on this site may earn us a commission at no cost to you. See how we review.
Send this guide
DollarFlourish Editorial
Data & Research Desk

The DollarFlourish Money Research Team builds the site's calculators and data rankings and writes its research-driven guides. Every figure we publish is traced to a primary source, the Bureau of Labor Statistics, Census Bureau, IRS, Social Security Administration, and Federal Reserve, and dated so you can check it yourself.

Reviewed for accuracy by Timothy E. Parker · Updated 2026-11-17 · Editorial & corrections policy

The Flourish Letter

One smart money idea each week, charts included. Join free and get the printable 2026 Money Calendar in your welcome email.