Top to bottom
Explore the page
- 01
How the stress test works
The page opens under the heading “How often this plan makes it” with three panels that show the method before any result.
The first draws every calendar year of the chosen index as a bar, with pills for the S&P 500, Nasdaq, Dow Jones and a 60/40 blend, an “S&P 500 → 60/40 · after retirement” preset, and a separate “Custom” control that assigns an index to each stretch of plan years.
The second shows one example deal: those years cut into stretches and laid in a new order. The third is the tally of endings. By default, the plan deals S&P 500 years while the household is working and switches to a rebalanced 60/40 portfolio in its first retired plan year; a household already retired starts in 60/40.
A plan with no account able to follow a market gets a note saying every run would be the same run, and nothing else.
- 02
Set up the test
Below the panels sit the controls the deal is made under: 100, 300 or 500 runs (300 is where the page opens), a “Deal it again” button that draws a fresh set of histories, and a “Deal settings” control whose summary reads as a sentence — for example “20-year stretches, no repeats · the dealt years’ own, year by year”.
While the workers are running, a progress bar counts the futures lived; once done, a line states how many runs were dealt, from which index and years, by which method, and at what inflation. An info button beside it explains the method in full.
- 03
Read the overall result
A row of figures reads the deal. “Runs that never went short” is the share of runs that made it, with “give or take” so many points beside it — half the width of a Wilson 95% interval, never less than one point.
Three smaller figures split the runs into those that kept every home, those that were driven to sell a home, and those that ran short; the last names the age by which half the short runs had already failed. “Middle run leaves, year N” is the median ending net worth. A bar below repeats the three shares as one strip.
During a What-If edit an extra line states the rate before and after, and how many runs changed their ending.
- 04
Explore the range of outcomes
The fan chart stacks every run’s net worth and reads across by year: a pale band from the 10th to the 90th percentile, a darker band from the 25th to the 75th, and the middle run as a line. The plan’s own flat-rate projection is drawn dashed for comparison, and during a What-If edit the pre-edit middle run is drawn underneath.
Hovering a year lists what each line stood at, roughly what share of runs stood above it, and how many runs had gone short by then. Percentiles are taken per year, so the band’s edge in one year is rarely the same run as in another. Pressing a run card or a table row overlays that run on the fan.
- 05
Review individual runs
Under the fan are cards for a handful of runs replayed in full. When some runs went short, the three that failed earliest are “The hardest run”, “Second hardest” and “Third hardest”; when none did, “The leanest run” is the one near the 10th percentile of ending worth. “The middle run” and “A fortunate run” sit near the 50th and 90th.
Each card names the calendar years dealt at retirement, the worst year lived, the inflation lived when it differs from the plan’s rate, when a home was sold or the run went short, and what it leaves behind.
“Every run” is a table of all of them — outcome, net worth at retirement, the annualized market return of the first five retired years, the largest net-worth drawdown, ending net worth — filterable by outcome, sortable, with an “Open” link that reads the whole projection as that run lived it.
- 06
Find when money runs short
Selecting a card or row opens two charts of that run: its net worth at its own scale, with the largest peak-to-trough decline shaded, the years it sold a home or went short dotted, and a strip beneath showing each year’s withdrawal as a share of the portfolio entering it (green at or under 4%, amber above, red past 6%); and the weather it was dealt, one bar per plan year with the worst year named and the end of working life marked.
When any run ran short, a final chart counts how many runs first went short in each plan year. The page closes by saying how many market-eligible accounts follow the index timeline and that the dice change nothing about the plan itself.
What changes what
Controls & settings
Check what each control changes and whether it recalculates your forecast.
Sets how many futures are dealt and lived. More runs narrow the give-or-take beside the headline; 300 is the default.
Advances the seed and deals a fresh set of histories to the same plan. A headline that holds still across deals has converged; a jumpy one says to run more.
Chooses the market the years are dealt from — S&P 500, Nasdaq, Dow Jones or 60/40 — or a timeline that switches index at chosen plan years, with each stretch dealt from its own index and no block crossing the boundary.
Picks the dealing method — long stretches with no repeats (the default, 20-year stretches), history restarted at random, one unbroken stretch, short blocks redrawn with replacement (5-year blocks by default), or every year at random — and the stretch or block length. Changes apply only on “Save & re-deal”.
Picks the inflation model: the dealt years’ own CPI in place (the default), the plan’s rate, a flat rate, one rate per run drawn evenly in a range, or one rate per run drawn from a bell around a center. This reprices the same futures without reshuffling which years any run is dealt.
Shows every figure at today’s purchasing power, discounting each run at the inflation that run itself lived, or in the dollars of each year.
Selecting a run shows that one run whole — its net worth, drawdown, withdrawal strip and dealt weather — and overlays it on the fan. Choosing a table row replays that single run; the cards were already replayed when the deal settled. The deal’s counts do not change.
Narrows the table to runs that kept homes, sold a home or ran short, and orders the rows by any column.
Reading the result
Read the results
The headline counts a run as making it when no year closed more than $0.50 below zero after every allowed withdrawal and sale; a run driven to sell a home is inside that rate and reported separately, because it is a different retirement to have lived.
The give-or-take measures only the shuffle: it says where the rate would settle with more deals from this same history. It says nothing about whether the plan’s own figures are right or whether the years ahead resemble the years behind.
The fan’s bands are read per year across all runs, not along one run; the 10th-percentile line in one year and the next belong to different futures, so no band edge is a path anyone would live.
The dealing method is the model’s own opinion: the default keeps 20-year stretches together and never deals a calendar year to the same run twice, which is deliberately harsher than drawing years one at a time. Switching methods shows how much of the rate is the sampling itself.
When runs go short, where they bunch matters more than the count: half of the short runs failing at 90 is a near miss, half failing at 70 is a different retirement.
One plan, several readings
Related pages
Projection →
The dashed line on the fan is the projection page’s own flat-rate run. Each row of Every run has an “Open” link that reads the whole projection — cash flow, taxes, net worth year by year — as that one run lived it, with the ordinary plan drawn underneath; the address carries the seed, run number, index, sampling and prices so the same run opens at the other end.
Projection →
Which accounts follow a market is set on the projection page, with the “Real market” control beside the charts. Accounts already following an index ride each dealt sequence as-is; accounts on a fixed rate keep their own long-run return and borrow only the ups and downs; cash accounts stay on their cash path.
Projection →
A What-If edit on the plan deals the identical futures to the plan the edit started from, so the line above the verdict bar states the difference the edit made — rate before and after, and which runs changed their ending — rather than the noise of two separate deals.
Keep following the math
Key concepts
Keep in mind
Model limits
No run is a forecast. Every run is the plan lived through one more ordering of years that really happened, and the rate is a count of those endings, not a probability about the future.
The page changes nothing about the plan: the index picker, dealing method, inflation model, seed and run count are settings of this page, and the dice leave the plan’s own returns, balances and rules as they are.
The rate and its interval depend on the chosen record, the dealing method and the run count; they do not cover errors in the plan’s own numbers, tax law, or risks absent from the historical series.
With no account able to follow a market, the page runs nothing and says so; it cannot stress-test a plan held entirely in cash and fixed rates.
The give-or-take is a Wilson 95% interval on the share of runs; it is not an interval on any dollar figure, and percentile bands are summaries of model runs rather than guarantees.
This guide documents the page as built. The figures on it follow the plan’s own assumptions; none of it is individualized tax, legal, Medicare, or investment advice.