How Runaway calculates your trajectories.

One projected year, step by step

The deterministic projection is no black box: it is the same sequence of six operations, repeated every year through the selected horizon.

• Opening balance — The value of each holding at the start of the year, in its native currency.
• Growth — Each asset compounds at its effective CAGR. No return is drawn at random at this step.
• Contributions — Monthly contributions compound within the year at the CAGR’s monthly-equivalent rate, so they earn growth the same year they are made.
• Forced flows — Required minimum distributions, scheduled conversions, and dated life events apply before any other decision.
• Withdrawal — The year’s spending need, expressed net, is grossed up to cover tax, then drawn from your accounts in your priority order.
• Closing balance — Whatever remains becomes next year’s opening balance. Each account’s cost basis is updated along the way.

Spending is entered in today’s money and then indexed to inflation. Tax-bracket thresholds are indexed the same way so they stay constant in real terms — without that indexing, inflation would artificially push every plan into higher brackets.

What taxation covers, country by country

Runaway models three tax residences — France, Canada, and the United States — using 2026 schedules for a single filer. Three mechanisms are common to all three.

• Grossing up withdrawals — A net need is converted into a gross amount that covers the tax due: in closed form for flat-rate accounts, by bisection for progressive schedules.
• Cost-basis ratio — In a taxable account, the taxable share of a withdrawal is the amount × (1 − cost basis ÷ value). Contributions raise the basis, a withdrawal scales it by the fraction sold, and market growth never raises it.
• Currency conversion — Schedules, legislated in local currency, are converted into the plan currency when it differs from the residence currency.

Tax-free wrappers stay exempt abroad only when a treaty recognises them: a US Roth does for a Canadian or French resident, a TFSA and a PEA do not.

Two complementary views

How Monte Carlo varies the markets

The engine replays the plan thousands of times. Your CAGR stays the input: only the dispersion around it is random.

• Geometric centering — Your CAGR is treated as the median compound return, not the arithmetic mean. Dispersion therefore spreads symmetrically around it.
• Mean reversion — Each year, the accumulated deviation from trend is pulled back by a coefficient κ = 0.15. A great decade does not extrapolate forever.
• Four shock models — Gaussian shocks; fat tails through a Student-t distribution with ν = 5 degrees of freedom; a crash-aware mode; or a historical cohort that replays real history in 8-year blocks to preserve the sequence of good and bad years.
• Correlations — Shocks across assets are correlated through a Cholesky factorisation of the correlation matrix: asset classes do not fall independently of one another.
• Crash regime — 6% of years switch into a crash regime: volatility multiplied by 1.7, an additional drift of −12%, and correlations among risky assets lifted toward a floor of 0.85. That is precisely when diversification stops protecting you.
• Annual bounds — A single asset’s yearly return is capped to the −95% / +200% range, so a distribution tail cannot produce absurd years.
• Success rate — The share of paths that fund the full horizon. By common financial-planning convention, at least 85% is considered strong, 70–85% borderline, and below 70% fragile.

Model parameters

These values are read straight from the engine when this page is built, so they cannot drift from the code that runs in the application.