BEHIND THE NUMBERS
Methodology
A useful result starts with clear assumptions.
Look back to explore what’s ahead.
FIRE is being built to evaluate retirement plans across historical market sequences. Each simulation will apply a plan to a different starting year, showing a range of historical outcomes.
The app uses a fixed 1928–2025 snapshot of Damodaran’s annual US stock and 10-year Treasury returns, paired with BLS CPI-U, not seasonally adjusted, retrieved through FRED. The dataset version is 2026-10-30.1. Stocks include dividends; the Treasury series uses a constant-ten-year-maturity repricing approximation, rather than the return on a single bond aging to nine years.
Data sources and timing
Annual nominal stock and bond returns come directly from the source workbook, without interpolation. Inflation uses January CPI of the following year divided by January CPI of the current year. These are monthly averages, not exact trading-day endpoints. October 2025 CPI is missing in the downloaded BLS series; it does not affect the January endpoints used here.
Data: Aswath Damodaran, NYU Stern; and U.S. Bureau of Labor Statistics, CPIAUCNS, retrieved from FRED, Federal Reserve Bank of St. Louis. Shiller and Macrohistory are not used in this released snapshot. This app does not claim numerical parity with FI Calc.
How a simulation year works
At the beginning of each year, the plan requests its annual withdrawal. Assets fund it proportionally to their current values, without borrowing. Returns are applied to the remaining assets, followed by fees, then year-end rebalancing.
You can set annual stock and bond fees, fixed cash growth, and a rebalancing interval or disable rebalancing. Fees are applied to each asset after its annual return. Optional allocation targets change evenly, slowly at first, or quickly at first; trades happen only at scheduled year ends toward the following retirement year’s target. Without rebalancing, holdings drift and withdrawals remain proportional. A one-year plan uses initial weights. Inflation adjustments chain each year’s CPI ratio, including across loop boundaries. Year-end real values use cumulative inflation through that year.
A path succeeds only when every requested withdrawal is funded, allowing a rounding tolerance of one millionth of a dollar. Ending the final year with zero assets is allowed if all withdrawals were funded. The first shortfall identifies the failure year. Income and extra withdrawals occur at the start of the year alongside the base withdrawal. Income offsets their combined amount; any surplus is deposited into cash before returns. A shortfall measures the remaining portfolio withdrawal that cannot be funded. Extra withdrawals never inflate subsequent base withdrawals. Constant Dollar uses the initial annual amount and its inflation option. Percent of Portfolio applies the selected percentage to the current starting balance each year. Basic 1/N divides that balance by remaining years, including the current year; it has no ending target or withdrawal floors/caps. Both variable strategies can fund very little spending and still show funding success. Income offsets and extra withdrawals apply after the base strategy is computed. Schedules use inclusive retirement-year numbers; an event-start inflation adjustment starts at 1 in its first scheduled year and accumulates annually thereafter, including between recurring payments.
VPW and Dynamic SWR
VPW uses the beginning-year age and actual holdings with an age-100 planning horizon, assumed real returns of 5% for stocks and 1.9% for bonds, and a 10% withdrawal cap. Cash uses 0% assumed real growth as a FIRE extension. It does not include worksheet pension bridges or withdrawal limits.
Dynamic SWR is an annual annuity-formula variant using editable nominal growth and inflation assumptions. Base requests are capped at the available balance, including in the final year. This cap is a FIRE extension. Historical returns still determine the simulated portfolio; expected returns do not replace market history.
Both allow spending to shrink and do not guarantee that living costs are covered. These implementations do not claim full equivalence with FI Calc.
Smoothed withdrawal strategies
The 95% Rule starts at the selected portfolio percentage, then requests the greater of that percentage of the current balance or 95% of the previous calculated base withdrawal. The floor is nominal and does not adjust for inflation.
Endowment (smoothed variant) starts at the selected percentage, then blends 80% of the previous base withdrawal adjusted by the prior year’s inflation with 20% of the selected percentage of the current balance. This variant omits Yale’s historical valuation lag, payout bands and taxes.
Both carry only the previous calculated base request, not extra expenses, income offsets or the funded amount. Smoothing can request more than the remaining balance, so it can cause a funding shortfall. Neither guarantees spending adequacy.
Vanguard Dynamic Spending
The first base withdrawal is the chosen percentage of the initial portfolio. Later years use that percentage of the current balance, bounded around the previous calculated base withdrawal adjusted by the prior year’s CPI change. Defaults allow a 2.5% real decrease or a 5% real increase per year; both are editable.
Income and extra expenses are applied afterward and never enter next year’s base-withdrawal history. These are annual change limits, not guaranteed minimum living expenses. The floor can exceed the remaining balance and cause a shortfall. No separate lifetime spending floor, taxes or minimum-dollar override is modeled.
Vanguard research and worked example ↗Guyton-Klinger spending rules
First-year spending is the selected percentage of the starting balance. Later requests start with the previous calculated base adjusted by annual inflation. Positive inflation increases are skipped after a negative net investment return when the implied withdrawal rate is above the initial rate; deflation is retained.
After that adjustment, an implied withdrawal rate more than 20% above the initial rate triggers a 10% cut, except during the final 15 retirement years. A rate more than 20% below the initial rate triggers a 10% increase. Equality does not trigger a change. The resulting base request becomes next year’s reference, independently of income and extra expenses.
This is a spending-rules variant. It retains this app’s proportional sales and chosen rebalance schedule, omitting the original paper’s asset-sale priority rules. Returns used to identify losing years exclude withdrawals and income deposits, and include asset fees. Requests can exceed the balance and fail.
Sensible Withdrawals
The base is the selected percentage of the initial portfolio, adjusted by cumulative inflation. Each later year also adds the selected share of the prior year’s positive real investment gain. The first year has no gain-based supplement.
Prior real gain, expressed in the next year’s nominal money, is the ending portfolio minus the amount invested after start-of-year cash flows multiplied by that year’s CPI ratio. Returns are net of fees; withdrawals and income deposits are not themselves treated as gains or losses. Negative real gains add no supplement.
Gain-based extras are separate from scheduled extra expenses. Neither the base nor the supplement is capped at the available portfolio; funding can fail. Starter rates of 3% base and 10% gains share are examples, not a guarantee of safety.
Hebeler Autopilot II — PMT variant
This variant calculates a zero-ending-value, beginning-year annuity payment using the current balance, assumed real return and remaining plan years. The first year uses that PMT amount. Later requests blend 75% of the prior calculated base adjusted by annual CPI with 25% of the fresh PMT amount.
The return assumption is editable from 0–20%; the 3% starter value is an example. It does not replace historical market returns. The horizon is your plan’s remaining years, not a mortality or IRS life-expectancy table. No ending-target control, minimum spending floor or balance cap is applied.
The 75/25 concept follows the reference description; the timing, horizon, initialization and inflation treatment are explicit FIRE choices. It is not a claim of full FI Calc or original Hebeler algorithm equivalence.
CAPE-based withdrawals
The annual request is current portfolio × (base percentage ÷ 100 + multiplier ÷ CAPE). For example, a 1% base and 0.5 multiplier with CAPE 20 requests 3.5% of the current balance. There is no separate inflation escalation or balance cap.
A local annual supplement must cover every selected market year. Each path uses January CAPE for its replayed market year, including after a loop. Supplements remain in this browser and plan backups; no Shiller observations are bundled with the app. January values from revised monthly data are an annual modeling convention, not a certified January 1 point-in-time trading signal.
The valuation signal is for US equities but applies to the entire portfolio in this strategy. Source redistribution review remains separate from local use; no full FI Calc parity or spending-adequacy guarantee is claimed.
Perspective comes with limits.
Historical outcomes do not predict future returns or guarantee that a plan will succeed. A high historical success rate is not a promise about your retirement.
This app is for educational exploration and does not provide personalized financial advice. Consider your circumstances and professional guidance before making financial decisions.
Inspired by FI Calc.
The product specification draws on FI Calc’s retirement backtesting approach. FIRE has not yet implemented or verified numerical equivalence with it.
Read the reference documentation