What Is a Good Monte Carlo Success Rate?
Quick answer: There is no single number that makes a Monte Carlo score good, and this page will not hand you one. A success rate is the share of simulated paths in which the money lasted, under the assumptions on screen. In our default example it was 87%, and it moved to 83% when only the data window changed. What counts as enough depends on how much leftover risk you can live with.
Education, not financial advice. This page explains a method and shows what a model produced from stated assumptions. It does not tell you what to buy, sell, save or spend. For your own situation, ask a qualified professional.
What the number is counting
Every score you will meet on a Monte Carlo simulation retirement calculator comes from the same short procedure. The tool builds thousands of invented market histories, runs your plan through each one, and reports the share that still had money at the final age. If 4,000 of 5,000 paths lasted, the tile says 80 percent, to take a round example.
That is a count of paths inside a model. It is not a measure of how likely your retirement is to work, and it is not a grade. The phrase "success" is also a little generous: it only means the balance stayed above zero until the plan-to age, with no regard to how close to zero it came, or to how uncomfortable the years in between would have been. If the idea of random paths is new to you, what a Monte Carlo simulation is builds one step by step.
The hypothetical thresholds people ask about
Readers often want a line: is 70 enough, or 80, or 90, or 95? As hypothetical examples, those four numbers describe four different appetites for leftover risk.
A reading near 70 would mean that in roughly three of every ten simulated paths the money ran out early. A reading near 80 means about two in ten ran out. Near 90, about one in ten, and near 95, about one in twenty. That is all they mean. Rules of thumb that attach adjectives to these levels are folklore as far as this site is concerned: we have not found a source we can read that fixes the cutoff, and a cutoff would depend on the person anyway.
What can be said is that the same count carries different weight for different households. Someone with a large pension and a house they would be willing to sell experiences a failed path differently from someone whose whole retirement rests on one portfolio. The count is the same, and the stakes are not. Only you can set your threshold, and a page like this one cannot do it for you.
Why a perfect score is not the goal
It is tempting to read the top of the scale as the target. In a model, though, a score close to 100 usually has a price, and the price shows up as money that never gets spent.
Look at the example plan again. In the middle of its simulated paths, the balance at the plan-to age was $2,507,000, in today's dollars, and at the 75th percentile it was $5,402,000. Yet at the 10th percentile the balance was $0. The paths that fail and the paths that finish with a very large amount are produced by the same assumptions. To push the failures toward zero you would have to cut spending in every path, including the paths that were going to end with millions unspent.
A simple withdrawal example makes the trade clear. Take a hypothetical $1,000,000 portfolio held 60% in stocks and 40% in bonds, drawn down for 30 years, using the 1971 to 2025 assumptions. Spending $30,000 a year gave 99%. Spending $40,000 a year gave 95%. Spending $50,000 gave 80%, and $60,000 gave 59%. In the example, living on ten thousand dollars a year less than the second rung buys only a few points of score at the top. Each step up the ladder after that gives up more points for the same ten thousand dollars of extra spending. Whether that is worth it is a judgment about your own life, which the model can describe but cannot make.
What this number leaves out. It comes from a model with the assumptions shown, in annual steps and after-inflation returns. It does not model taxes, health or long-term care costs, inflation that differs from the model, or how you would act in a bad year. See what a Monte Carlo simulation cannot see.
Why two calculators disagree about the same plan
Most people who run a plan through two tools get two answers and wonder which tool is broken. Often neither is. The scores differ because the tools make different choices, and those choices are large.
Start with the data window. Our calculator's default uses real returns from 1971 to 2025, while the alternative uses 1928 to 2025. The example plan scored 87% on the first and 83% on the second, with every other input identical. Neither window is wrong. They describe different stretches of history, and the second contains decades that the first does not.
Next come the return assumptions. Change only the assumed average stock return and the count shifts a long way. With the stock return lowered by one percentage point a year, the example plan scored 80%; lowered by two, 69%; lowered by three, 57%. Raised by one, 93%. A tool whose authors chose a cautious return will report a lower score than one whose authors chose an optimistic return, even though the engines are identical. Volatility matters in the other direction: adding three points of annual volatility gave 82% and removing three gave 93%. A lower assumed bond return, by one point, gave 82%.
Then there is random noise. Run the same inputs with eight different seeds and 1,000 trials gave 87% to 89%, 5,000 trials gave 87% to 89%, and 10,000 trials gave 87% to 88%. A gap between two tools smaller than that spread tells you nothing. A gap of several points tells you the assumptions differ, and it is worth finding out which. The ten checks on the page about how to pick a Monte Carlo retirement calculator are a quick way to compare.
Differences also come from what a tool leaves out. Some tools model taxes. Some assume inflation is a steady figure and some draw it at random. Some treat Social Security as a fixed stream and some let you adjust it. All of these fall under how a model is built, and the full list of this site's choices is on how this simulation works.
How much the score moves when one thing changes
The table below shows the same example plan, with one assumption changed at a time. It is the best answer we can give to the question of how stable a score is.
| Change | Success rate |
|---|---|
| Default example plan | 87% |
| Retire 5 years earlier | 57% |
| Retire 5 years later | 98% |
| Spend 10% more in retirement | 81% |
| Spend 20% more in retirement | 74% |
| No Social Security or pension income | 64% |
| Plan to age 100 | 84% |
| Plan to age 105 | 83% |
| Stock return 1 point lower | 80% |
| Stock return 2 points lower | 69% |
| Stock return 3 points lower | 57% |
| Stock return 1 point higher | 93% |
| Stock volatility 3 points higher | 82% |
| Stock volatility 3 points lower | 93% |
| Bond return 1 point lower | 82% |
What this number leaves out. It comes from a model with the assumptions shown, in annual steps and after-inflation returns. It does not model taxes, health or long-term care costs, inflation that differs from the model, or how you would act in a bad year. See what a Monte Carlo simulation cannot see.
Read it for its pattern, not for any single cell. The score is sturdy to some changes and sensitive to others. If a small tweak to a return assumption swings the answer by many points, the headline number was never as firm as its digits suggested. This is why a result from one tool, run once, deserves less weight than a result you have probed by changing the assumptions yourself.
A second opinion from history
Because a score depends so much on how the model is built, it helps to see it beside something built differently. A historical replay does not draw random numbers. It runs a plan through actual sequences. Using the same dataset, a 30-year plan at the 4% rule, a rule of thumb that spends a fixed share of the starting balance, succeeded in 65 of 69 starting years from 1928 onward, or 94%. Under the 1971 to 2025 assumptions, the Monte Carlo count for the same rule was 95%, and under the 1928 to 2025 data it was 92%. Those are three numbers for one question, produced by three methods, and they agree to within a few points.
The caution is that the 69 historical windows overlap, so they are far fewer independent experiences than the count suggests. The 4% rule page and the Monte Carlo vs historical simulation page both go into this. Seeing agreement between methods is reassuring about the arithmetic. It cannot make the future resemble the past.
A score means different things at different horizons
One more source of confusion is the length of the plan. Lasting 20 years and lasting 40 years are different problems. In the withdrawal example above, the 4.0% rung scored 100% over 20 years, 95% over 30, and 87% over 40. The spending was identical; only the horizon changed. So a score quoted without its horizon is hard to interpret. If you are comparing your result with a friend's or with something you read, check that the two plans end at the same age. The question of how long savings last at different spending levels gets its own treatment in how long your money will last in retirement.
Four questions to ask before you trust any score
Instead of asking whether a number is good, it can help to ask what stands behind it. First, what returns did the tool assume, and are they after inflation? A score built on cheerful returns is a statement about the returns. Second, how long does the plan run? A plan to 85 and a plan to 100 are different tests, as the horizon rows above showed. Third, what is being left out: taxes, health costs, shocks to spending, a spouse who may live longer? Fourth, how much does the number move if you nudge the assumptions? If a modest change swings it by ten points, then you have learned that the score is fragile, which is useful to know in itself.
None of these questions produces a verdict. They help you see how much weight one figure can carry, and they turn a lone number into something you can compare with other evidence.
What this can't tell you
This page can't tell you what score to aim for, whether your own number is acceptable, or whether your plan will work. It cannot tell you your spending flexibility, your other assets, your health, or what the markets will do. It does not know your tolerance for a bad outcome, and that is the main thing a threshold should reflect. The figures here come from the default example inputs and the stated data windows, and they are illustrations of how a model behaves. If you want to run your own numbers, the retirement Monte Carlo calculator on the home page shows every assumption beside the result.
Frequently asked questions
Is 80 percent a good Monte Carlo success rate?
This page does not rate scores as good or bad. Taken as a hypothetical example, 80 percent means that in about eight of ten simulated paths the money lasted to the plan-to age under the assumptions shown. Whether that is acceptable depends on your income sources, flexibility and comfort with risk, which a model cannot know.
Why not aim for the top of the scale?
In a simulation, pushing the score toward the top usually means lowering planned spending, and that often leaves large unspent balances in the paths that were going well. In the default example, the median path finished with $2,507,000 left. Each household weighs that trade differently, so there is no universal answer.
Can two calculators both be right with different results?
Yes. They can differ in data window, return assumptions, volatility, trial counts, inflation treatment and what they leave out. Using the same plan, our calculator gave 87% on 1971 to 2025 data and 83% on 1928 to 2025 data, so a gap of that size can come from the data alone.
Does a score of 90 percent mean a one in ten chance of failure?
Not in the sense of a forecast. For example, it means one in ten of the simulated paths ran out of money before the plan-to age, given the assumptions. If the assumptions are off, or reality contains events the model lacks, the real frequency could be different. The count describes the model.