Projections & Monte Carlo
Every plan produces two results at once. The detailed projection runs the plan along a single return path and reports every year in full: taxes, cash flow, and balances. The Monte Carlo runs the same plan a thousand times against real slices of market history. It reports the range of outcomes: how much spending each run sustained, what it left behind, and what fraction of the runs succeeded.
These are not two settings you choose between. FinZot always computes both. The detailed projection is the run you read year by year. The Monte Carlo is the summary that runs alongside it. Both use the same plan and the same yearly steps (How a Year Works). Only the source of each year’s returns and inflation differs.
The detailed projection
Section titled “The detailed projection”The detailed projection is the single run reported in full. Every detailed number in this reference comes from it. That includes the year-by-year tax tables, IRMAA tiers, cash flow, and account balances.
By default it uses your expected returns. Each asset earns its long-run historical average return, the same every year. If you have given a fund a return-rate override, it earns that rate instead (asset return overrides). You can switch the source to one of two others. Each is still a single path shown in full.
- Historical replay. A named window of real history, such as the Great Depression or the 2007 to 2009 financial crisis, is spliced in starting at a point in the plan you choose. When the window ends, it either loops or reverts to your expected returns. This is how you ask “what if 2008 hit right at retirement?” A replay uses those years’ actual returns, so your return-rate overrides do not apply to it.
- Single trial. Any one Monte Carlo trial replayed in full detail, so a specific sampled history becomes visible year by year.
How a return-rate override reaches each source is set out in asset return overrides. It applies on the expected-returns path, is carried into a replayed trial after recentering, and is dropped by a historical replay. Market data in each run maps every market data input to the run it reaches.
Whichever source is active, the detailed projection is one path. It answers how and why a plan behaves. It can never answer how often a plan survives, because no single path can.
The Monte Carlo
Section titled “The Monte Carlo”A single path cannot show sequence-of-returns risk. A bad decade early in retirement sinks a plan that the same decade late in retirement survives. An averaged path has no sequence, and a historical replay is only one sequence.
So FinZot always also runs the plan as independent trials, 1,000 by default. Each trial builds its returns with a block bootstrap. It takes consecutive slices of real history, 10-year blocks by default, and stitches them together to the plan’s length. This preserves real runs of good and bad years in order. Inflation is sampled the same way, so each trial carries its own path of today’s dollars. You can change the trial count, block size, and random seed. The same seed always produces the same trials. These settings are plan-wide, shared by every scenario (scenarios). How each trial is built, and rescaled to your own return assumptions, is explained in Scaled Block Bootstrap.
Every trial is a complete plan run, so FinZot reports the whole range: the spending each trial sustained, the net worth it ended with, when its milestones were hit, and how it finished. The success rate is the share of trials that succeeded. If 850 of 1,000 trials succeeded, the success rate is 85%.
What “success” means
Section titled “What “success” means”Every trial ends in one of three verdicts, named after the two classic insolvency tests. It is solvent when it funds every year and ends with net worth at or above zero. It is cash-flow insolvent when a year could not be funded from cash plus securities. It is balance-sheet insolvent when every year was funded but the plan finishes owing more than it owns. The success rate counts solvent trials only. FinZot reports the number of balance-sheet insolvent trials beside the rate whenever any trial ends that way.
A year fails the moment the plan cannot cover that year’s required spending and taxes from cash plus the sale of its investments. At that point the liquid assets are gone with nothing left to draw on (allocation, lots, and selling). The trial is cash-flow insolvent and the projection stops at that year. There is no recovery.
The ending test runs once, on the plan’s last year. Net worth there is cash plus investments plus property at market value, minus every loan and mortgage balance. A plan that finishes below zero is balance-sheet insolvent. The detailed projection marks its last year on the timeline when it does.
What counts as required spending depends on whether Variable Spending is on.
- Off, the default. Every expense you entered is paid in full each year. The bar is the whole plan: all expenses, plus taxes, debt payments, and property costs.
- On. A spending policy sets each year’s budget from the portfolio, and that budget flexes. A bad decade lowers spending instead of exhausting the plan. What it can never cut below is your spending floor: expenses marked mandatory, plus the Spending Floor percentage kept on each flexible expense. Debt payments and property costs sit outside the budget and are always paid in full. That floor is the bar, so the success rate measures the floor, not the budget.
The two rates are therefore not comparable. A variable spending plan reporting 95% held its floor and finished above water in 95% of trials, while spending well above the floor in most of them. It did not sustain the spending you entered 95% of the time. Its risk shows up as a lower sustained-spending range, so read the two together.
Funding a year is a liquidity test, not a net worth test. The ending test is the one place net worth enters.
- Only spendable wealth funds a year. Home equity and other illiquid assets neither count toward staying afloat nor get sold to rescue a year. A plan can fail while still owning a paid-off house.
- Debt is invisible until the end. A loan is paid every year and never fails a year on its own. Its balance is subtracted only in the ending test. So a loan whose principal never comes due inside the plan, such as a revolving line or a balloon dated after the last year, can leave a fully funded plan balance-sheet insolvent.
- Failure is all or nothing. A trial ending with $1 of net worth counts as a success. The rate counts survivors, not how comfortably they survived. It is one summary of the trials, not the whole picture. The sustained-spending and ending net worth ranges are reported beside it, and they separate plans the rate alone cannot.
The two results side by side
Section titled “The two results side by side”| Detailed projection | Monte Carlo trials | |
|---|---|---|
| How many runs | One | 1,000 by default |
| Returns each year | One path: expected returns by default, or a historical replay, or a single trial | A different real slice of history per trial, from the block bootstrap |
| What it reports | Full year-by-year detail: taxes, cash flow, balances | A full run per trial, summarized as ranges of spending and ending net worth, plus the success rate |
| Question it answers | How and why a plan behaves | How often, and how well, the plan holds up |