Use Your Own Capital Market Assumptions in a Monte Carlo Retirement Plan
Published capital market assumptions disagree with each other — sometimes by two or three percentage points on expected equity returns. If you are an advisor with a house view, or a hands-on investor who has formed your own opinion about the next decade, you may not want to adopt one published source and live with it. You want to enter your own numbers and watch what they do to the plan.
QuantCalc now supports custom capital market assumptions on the free Monte Carlo simulator. You can type in your own expected returns, volatilities, and the full correlation matrix across the five asset classes, then run 10,000 paths against them.
Enter your own assumptions — FREE
Pick “Custom (my assumptions)” in the Market Assumptions panel, edit the returns, volatilities, and correlation grid, and run. No account needed.
Open QuantCalc →What you can change
The simulator models five asset classes: US stocks, international stocks, bonds, real estate, and cash. For each one you can set:
- Expected return — your annual forecast for the asset class.
- Volatility — the annual standard deviation of returns.
- Correlations — the full pairwise grid that controls how the assets move together. This is the input most calculators hide from you, and it is the one that drives tail risk.
Your assumptions are saved in your browser, not in an account, and are sent only to run each simulation. Switch back to a published source any time — J.P. Morgan, BlackRock, Vanguard, GMO, Schwab, or Invesco — to compare your view against the institutions.
The part most tools get wrong: invalid correlation matrices
Here is the problem nobody warns you about. When you type correlations by hand — say US stocks and real estate at 0.8, real estate and bonds at −0.7, bonds and US stocks at 0.6 — the grid you produce is very often not a valid correlation matrix. Mathematically it has to be positive semi-definite; in plain terms, the relationships have to be mutually consistent. You cannot have three assets that are each strongly correlated in a way that contradicts itself, any more than you can have three people who are each two inches taller than the other two.
An invalid matrix breaks the math that generates correlated random returns (the Cholesky decomposition). Most tools do one of three unhelpful things: reject your input with a cryptic error, silently fall back to treating the assets as independent (which understates risk), or produce garbage without telling you.
QuantCalc does something different. If the matrix you entered isn't valid, it repairs it to the nearest valid correlation matrix, runs the simulation on the corrected version, and shows you a notice saying it adjusted your input — including the largest change it had to make. You stay in control, and the result stays mathematically sound.
“Nearest” here is precise: the engine finds the smallest set of adjustments (by eigenvalue clipping) that makes your grid consistent while keeping every asset's variance fixed. If the largest change it reports is tiny — say 0.02 — your assumptions were almost valid and you can trust the run. If it is large, that is a signal your correlation views are internally contradictory and worth revisiting.
Geometric vs arithmetic returns
One technical detail decides whether a long-horizon plan is honest. A geometric (compound) return is lower than the arithmetic average of the same return stream, and the gap grows with volatility. Feed an arithmetic average into a model that compounds it and you overstate growth over a 30-year retirement. Custom assumptions in QuantCalc default to the geometric convention, the same convention the forward-looking published sources use, so the median multi-year growth matches the return you actually typed. (For more on why this matters, see forward-looking vs historical CMAs.)
A practical workflow
- Start from a published source you respect so the volatilities and correlations are already sensible.
- Switch to Custom and change only the inputs you have a real view on — often just the equity return.
- Run it and read the success rate. Then compare against the unmodified published source to isolate the effect of your view.
- Stress it. Lower your equity return by two points and re-run. If the plan falls apart, your plan was leaning on an optimistic assumption, not on robustness.
This is the same discipline an advisor uses when presenting a range rather than a single number: the point isn't to find the “right” forecast, it's to learn how sensitive your plan is to the forecast being wrong.
Who this is for
- Advisors who maintain a house capital market view and want to run client plans against it rather than a vendor's.
- DIY investors who think published equity forecasts are too high (or too low) and want to see the consequence of their own number.
- Anyone stress-testing a plan against a deliberately pessimistic decade — low returns, high correlations — to find the breaking point.
Frequently asked questions
Is the custom-assumptions feature free?
Yes. Entering your own returns, volatilities, and correlations and running the Monte Carlo simulation is free, no account required. The portfolio optimizer and multi-source forecast comparison remain PRO features.
Do my assumptions get stored on your servers?
No. They are saved locally in your browser and transmitted only to run each simulation. Clear them any time by switching back to a published source.
What if I enter a correlation matrix that doesn't make sense?
The simulation still runs. QuantCalc repairs the matrix to the nearest valid one, uses the corrected version, and tells you what it changed so the result is never silently wrong.
How many assets can I customize?
All five modeled classes — US stocks, international stocks, bonds, real estate, and cash — including every pairwise correlation between them.
QuantCalc is an independent educational tool, not affiliated with, endorsed by, or sponsored by any referenced firm. Not financial advice.