# Design notes

Why Escape Velocity is shaped the way it is: what it refuses to model, which
number it puts first, and where its honesty budget is spent.

## What this is, and what it isn't

Every serious FIRE calculator is better than this one at the maths.
[FIRECalc](https://firecalc.com) backtests your plan against every historical
sequence since 1871. [cFIREsim](https://cfiresim.com) adds variable withdrawal
strategies. [ProjectionLab](https://projectionlab.com) does tax-aware full-plan
modelling with Roth conversions and state brackets. This runs one deterministic
projection at a fixed real return and models no sequence risk, no tax, no asset
allocation, no pensions, no multiple accounts.

**And you open them once a year.**

That's the whole argument. A calculator you visit each January is a thing you
consult; a number sitting in your status bar is a thing you *live beside*. You
see it while you work, forty times a day, on the days you were not thinking
about money at all — which are exactly the days the decisions get made. Not the
retirement decision. The subscription, the upgrade, the second monitor, the
job offer with the higher salary and the worse hours.

This is why everything is priced in days rather than currency. A €40/month
subscription is not €480 a year, it is nearly **six days of your life every year,
forever**. You cannot hold that thought from an annual spreadsheet visit. You
can hold it when the number is on the screen while you are deciding.

So: worse instrument, better placement. Work the real numbers out somewhere
serious, bring them back here, and let them sit where you can see them.

Two limitations worth stating plainly, since the above is not an excuse for
them. The 4% rule exists *because* sequence risk dominates early retirement —
this uses that conclusion without modelling the risk behind it, so treat any
date as a midpoint rather than a promise. And tax is ignored entirely, which is
not a rounding error everywhere: Irish investors, for one, face deemed disposal
on ETFs every eight years whether or not they sell.


## Money or time

The same rung reads very differently depending on the question:

| | money | time |
|---|---|---|
| Lean FI | 36.1% | **52.2%** |
| Full FI | 17.6% | **36.5%** |

Money is `balance / target` — checkable, and pessimistic, because compounding
means the last stretch arrives far faster than the first. Time is the share of
the journey already elapsed. Neither is wrong; **Progress shown as** in the
settings picks which gets the big type, and the other is always shown beside it
so no single framing gets to stand unchallenged.

**Time is the default.** Money is the more conservative claim — you can verify
`balance / target` without trusting the model at all — but it consistently
understates where you actually are, and the gap widens the further out the rung
is. Both are always on screen, so the check is one glance away.


## Honest by default

- **Real returns, not nominal.** The default is 5% *after inflation*, roughly 8%
  nominal minus 3%. Using the nominal figure makes every date on the ladder
  arrive years early and is the most common way FIRE spreadsheets lie to their
  owners.
- **Projections are labelled as projections.** Nothing here is a forecast of the
  market. It's compound interest applied to assumptions you chose, and it is
  exactly as good as those assumptions.
- **The 4% rule has limits.** It comes from the Trinity study on US
  equity/bond portfolios over 30-year retirements. It is not a law of nature, it
  performs differently for early retirees with 50-year horizons, and it does not
  transfer to a portfolio with a materially different volatility profile. The
  withdrawal rate is a setting for a reason.


## Definitions

| Rung | |
|---|---|
| **Runway** | months of expenses — the survival floor |
| **Walk-away money** | years of expenses; enough to leave without a plan |
| **Coast FI** | stop contributing today, still hit FI by your target age |
| **Lean FI** | pared-back expenses ÷ withdrawal rate |
| **Full FI** | annual expenses ÷ withdrawal rate (4% → 25x) |
| **Crossover** | the point your returns out-earn your contributions |

The two numbers in that block are the point:

- **you buy** — your monthly saving over the daily cost of living. The future
  days each month's saving pays for.
- **the pot earns** — the same, for growth the pot is *expected* to produce
  (`balance × r ÷ 12`). Expected, not last month's actual return: a real month
  on a six-figure balance swings by several times a contribution, so a realised
  comparison would flip back and forth for years and mean nothing.

When the second passes the first, your money is doing more of the work than you
are. That's the **crossover**, and it arrives quietly, years before any of the
rungs — no other number on the panel announces it.

The settings strip carries a third day-figure: how many days of every month the
pot already *covers* at your withdrawal rate. That one is the FI percentage in
different units (`balance × swr / expenses` is identical to `balance / fiNumber`),
with a finish line you can feel — at **30.4** the pot covers every day of the
month and you are done.

**Balance means invested balance.** The projection compounds it and the
withdrawal maths draws from it, so property equity, pensions you cannot access,
and cash beyond your runway will quietly move every date on the ladder if you
include them.


## It updates itself

You record a balance whenever you feel like it. Between those points the widget
projects forward daily, prorating your contribution across the days, so the bar
moves on its own and the milestone dates shift as you go.

When you do enter a fresh balance, it re-anchors the projection and stores a
point in your history — which gives you the thing spreadsheets don't:

> **ahead of plan by €2,098**

Actual against projected, so the model corrects itself instead of drifting.

If a balance goes unrecorded past 45 days the panel says so and the bar dims. A
confidently-displayed six-month-old extrapolation is a lie, and this refuses to
tell it.