Success odds

Each year of a trial takes three rates from one historical year: the U.S. stock total return, the U.S. Treasury bond return and consumer price inflation. The stock return is split into dividends and price growth using that year’s recorded dividend yield, so dividend income varies too.

These follow the sampled rates in every trial:

• Stock and bond growth in every investment account, blended by its bond allocation.
• Inflation of your expenses and today’s-dollar amounts.
• Raises on pay and Social Security increases with a blank growth rate.
• Federal brackets, deductions and limits after 2026.
• Anything that reacts to those results: taxes, withdrawals, flexible-spending cuts, and milestones triggered by net worth.

These stay fixed in every trial:

• Life expectancy, milestone ages and dates, and every entered amount and rate.
• Bond allocation, including a curve.
• Home and property appreciation, loan rates, and a medical inflation rate you set.
• An account’s own fixed stock growth and dividend yield. Only that account’s bond share varies.
• Tax law as modeled.

By default a trial strings together blocks of consecutive historical years, each block 3 to 7 years long and starting at a random year from 1928 to 2024. Blocks keep real runs of good and bad years, and the three rates in a year always come from the same historical year. Other sampling methods are described in the next section.

By default each sampled rate is then calibrated to your plan. Every year of the history is shifted by the same factor so that its long-run compound rate equals your assumption for that year. Year-to-year swings keep their historical size and order. For stocks, the target is your price growth plus your dividend yield.

Turn calibration off to use unadjusted history. Its compound stock return is about 10% a year, well above the 7% default, so the success rate usually rises. The example plan below goes from 69.1% calibrated to 90.9% uncalibrated.

Sampling method chooses how historical years are lined up in a trial. Every method uses the 97 years from 1928 to 2024, takes all rates in a year from the same historical year, and is calibrated to your plan unless you turn calibration off.

Return sources let you replace any of stock total return, stock dividend yield, bond return and inflation with a normal distribution you set by mean and standard deviation, in percent. A normal variable is drawn fresh each year, independently of the other variables and of history; the historical variables still come from the sampled year together. Draws are limited to a −100% return and a yield of zero or more, and calibration doesn’t adjust them. When all four are normal, the sampling method has no effect.

A normal mean is an average yearly return, not a compound rate. Ups and downs lower compound growth, so a 10% mean with an 18% standard deviation compounds at about 8.5% a year. Your plan’s assumed rates are compound rates, so a normal mean equal to your assumption grows more slowly than the steady forecast. Set the mean higher if you want the two to match on average.

Separate crypto return distribution gives cryptocurrency accounts their own yearly normal return in place of the stock price return, even when the account sets its own fixed growth rate. Crypto pays no dividends, so the draw is the whole return. Without it, crypto accounts grow like the stock price return.

Each trial still reruns exactly from the run’s seed and the trial number, whichever method you choose. For historical sequences, a rerun trial shows the historical year it started from.

A trial fails in the first year that either happens:

• Spending can’t be fully paid from the withdrawal order.
• The failure test is met at year end. The default test is that cash and investment balances total zero or less. The net-worth test uses total net worth instead, which adds property and subtracts debt.

The funding success rate counts every trial that never fails. Two optional targets narrow success further and appear as a separate target success rate:

• Final net worth target: the trial must end at or above this amount, in dollars of the final year. It starts at the plan’s Leave to heirs amount from Partner and estate, inflated to that year.
• Count flexible-spending cuts as a shortfall: a trial whose flexible-spending rules cut spending in any year doesn’t count as meeting targets.

Each trial lands in one outcome category: met all targets, survived with spending cuts, survived below target, or funding failure. Failure analysis groups failed trials by cause with their average age at failure.

Outcome detail adds two breakdowns. Failed trials are early, mid or late by the share of plan years completed before the failure year: early before a third of the plan, late from two thirds, by default. Funded trials compare final net worth with net worth at the end of the year your retirement milestone happens, or of the first plan year when there isn’t one or it never happens. Choose that milestone under Retirement milestone; left on automatic, it is the first milestone whose name mentions retirement and not a partner. Both amounts are in today’s dollars of that trial’s own inflation. Below 25% is a thin margin, above 200% a wide margin, and anything between a moderate margin. You can change all four cut-offs and the milestone under Outcome detail in the run settings.

Extra percentile charts add the same percentile bands as the net worth chart for net worth without real assets, spending, taxes or income, in nominal dollars.

The chart ranks all trials’ net worth separately in each year. The median line is the middle value in each year, not one trial’s path, and a trial can sit near the top one year and near the bottom later. The 10th percentile is the value that 10% of trials fall below. Amounts are nominal dollars of each year.

The median ending above or below the steady projection is normal. See Why the projection and success odds differ.

The percentile chart can’t show what one bad but funded trial looks like, or which trials reached a milestone late. The trial explorer below it lists every trial on its own.

• Histogram: each bar counts the trials whose final net worth falls in its range, with failed trials at the bottom in red. Select a bar to list only those trials, and select it again to clear it. The two outer bars also hold the few most extreme trials, so the rest stay readable.
• Outcome: all trials, funded, failed or, when a target is set, one outcome category.
• Milestone timing: trials that reached a milestone by an age, after an age or never. Ages are yours in the year the milestone happened.
• Order: by trial number, or by final net worth from lowest or highest.

Each row shows the trial’s outcome, final net worth, failure year and a small net worth line. Select a row to rerun that trial. It uses the same market history as in the run, so the result matches the row exactly. You see its net worth for every year against the median of all trials, plus its key years: peak net worth, the largest one-year fall, milestones reached, the failure year and the final year.

• With historical sources, returns and inflation outside the range of U.S. history since 1928, or any history other than U.S. stocks, Treasury bonds and consumer prices.
• Links between normal variables. Each normal draw is independent, so a bad stock year is no more likely to come with high inflation.
• Uncertain lifespans. Every trial ends at the plan’s life expectancy.
• Changes in tax law, benefit rules or your own behavior beyond the flexible-spending rules you set.
• Swings within a year. Each year gets one return, and failure is checked at year end.

To replay one specific historical period instead of random blocks, use the Historical stress test.