Deorbit Plan

Methodology

How the simulator works, what it leaves out and how it is tested · By the Deorbit Plan editorial team, a Berrydyne program

What the simulator does

A single “retirement number” can’t tell you how a plan holds up when markets, inflation and taxes don’t follow their averages. Deorbit Plan simulates your household year by year, thousands of times, each time with a different market history. It reports how often the money lasts and what taxes and premiums the plan pays along the way. Use it to compare choices (when to convert to Roth, when to claim Social Security, how to order withdrawals). It does not forecast your future.

Monte Carlo over market paths

Each simulated lifetime is one “path”. The success rate is the share of paths in which the money never ran out before the end of the plan. You can choose between two ways of generating markets:

  • Parametric (default). Stock and bond returns and inflation are drawn from bell-curve distributions whose averages, volatilities and correlation you can edit.
  • Historical bootstrap. Each simulated year is a real calendar year drawn at random from the historical dataset, so the relationships between stocks, bonds, cash and inflation within a year are real. Years are shuffled independently, so multi-year patterns such as long bear markets are not reproduced.

A fixed random seed makes a run repeatable. The year-by-year charts come from a separate run that uses average returns every year. For the details and default values, see How the simulation works in Help.

The annual loop

For every year of every path, the simulator does the same things in the same order:

  1. Grow each account with that year’s market returns.
  2. Add contributions and wages while anyone in the household still works.
  3. Take required minimum distributions, inherited-account distributions and Social Security.
  4. Make any Roth conversions the strategy calls for.
  5. Withdraw what spending, premiums and taxes require, following the chosen order.
  6. Compute taxes and premiums, and repeat the withdrawal-and-tax step until the two agree.

The last step repeats because a withdrawal can itself be taxable income, which raises the tax bill, which calls for a larger withdrawal. The loop runs until the two agree.

What is modeled

  • Federal income tax by filing status, with capital-gains stacking, the net investment income tax, the standard deduction, the additional deduction at 65 and the 2025–2028 senior deduction.
  • State income tax for a set of states, as simplified rate and deduction tables with each state’s Social Security and retirement-income rules.
  • Social Security: claiming age, benefit taxation, and survivor benefits in the survivor what-if.
  • Medicare Part B and Part D premiums and IRMAA surcharges (based on income from two years earlier), and ACA premium subsidies before 65.
  • Retirement accounts: pre-tax, Roth, HSA and taxable accounts, required minimum distributions, qualified charitable distributions, Roth conversions, and inherited accounts under the 10-year rule.
  • Life events: pensions, part-time work, lump-sum expenses, gifts, inheritances and a survivor scenario.

Every law-derived number is a sourced constant (see the editorial policy), and the 58 Learn articles explain the rules behind them.

Documented limitations

These are the main simplifications. The full list is under Assumptions & limitations in Help.

  • No alternative minimum tax and no itemized deductions; years with very high taxable income are flagged for extra caution.
  • State taxes are approximations: credits, local taxes and phase-outs are ignored.
  • Future tax law is assumed to stay as it is today.
  • The Social Security earnings test is not modeled.
  • One filing status applies throughout, except in the survivor what-if.
  • Mortality is not simulated inside each path; the charts show it as an overlay.
  • Taxable-account sales use an average-basis approximation, and tax-loss harvesting is not modeled.
  • The historical bootstrap preserves each year’s relationships but not multi-year sequences.

The disclaimer explains what these limits mean for how to use a result.

How the math is tested

  • Golden tax cases. Tax calculations are checked against cases worked out by hand from the published rules. Some are deliberately tricky cases written before anyone looked at the code. Separate sets of hand-worked cases cover filing statuses, the senior deductions and the survivor transition.
  • Cash-flow conservation. For every demo household that ships with the site, a test checks that in every simulated year what comes in (withdrawals and guaranteed income) equals what goes out (spending, gifts, premiums and taxes) plus any surplus held as cash.
  • Snapshot tests. A saved, rounded copy of one fixed run, plus checks on the order of the steps, catches any change to results that nobody intended.
  • The full suite. There are well over a thousand automated checks. The type check, the full suite and a clean build must all pass before any change goes onto the site.
  • Sourced constants. Each law-derived number records its primary source and whether that source has been checked. See the review process and the full list on the sources and verification page.

These tests show that the code does what it was designed to do. They can’t make the model a prediction of your future.

Privacy

The simulation runs in your browser, in a background process on your own device. There is no account, and no server receives your inputs. Your plan is saved only in your browser’s local storage. The privacy policy has the details, including the one exception (the optional feedback form).