Your FIRE Number Isn't Fake. Your FIRE Date Is.

Every few weeks someone posts a version of the same question: the Shiller CAPE ratio is sitting at 41.98 as of 26 August 2026, against a historical median near 16 and an all-time high of 44.19 set in December 1999. So is my FIRE number real, or is it just a screenshot of a bubble?

The number is real. It is 25 times your spending, and a bear market does not change what groceries cost. Nothing about a high market makes $2 million a wrong target for an $80,000 lifestyle.

What is fragile is your date. So I ran the stress test through Stoke's own projection engine. A 30% crash is not the main threat to a mid-journey saver. The bigger threat is the scenario that never makes the news: a decade of ordinary-looking 3% returns. And once you hold the ten-year total return constant, taking the loss now is better than grinding, right up until about five years before your FI date.

The number is denominated in spending, not in prices

This part is worth being precise about, because it disposes of the panic version of the question.

Your FIRE number comes out of your withdrawal rate and your annual spend. At a 4% rate, $80,000 of spending needs $2,000,000. Neither input is a market price. If the S&P 500 halves tomorrow, your spending is unchanged, your withdrawal rate is unchanged, and your target is unchanged. The 25x rule breaks in three specific ways, but "stocks are expensive" is not one of them.

What a crash changes is the balance on the left side of that comparison, and therefore the number of months until the two sides meet. That is the thing to stress-test.

What a 30% crash actually costs, in months

Take an illustrative household: 38 years old, $80,000 of annual spending, $2,000,000 target at a 4% withdrawal rate, saving $4,000 a month. Everything below is in today's dollars. The return figures are real returns, so inflation is already netted out, and contributions are assumed to hold their real value.

I ran this through run_projection in backend/src/services/projection.py, the same code path that produces the FI date on Stoke's planning surface, applying the crash as a one-time principal adjustment at month zero and reading back the month the projection first crosses the target. The baseline uses Stoke's default 7% real return.

Portfolio	% of target	Baseline to FI	After −20%	After −30%	After −40%
$200,000	10%	194 mo (16.2 yr)	+8 mo	+13 mo	+17 mo
$500,000	25%	144 mo (12.0 yr)	+15 mo	+23 mo	+32 mo
$1,000,000	50%	82 mo (6.8 yr)	+22 mo	+35 mo	+48 mo
$1,500,000	75%	37 mo (3.1 yr)	+26 mo	+40 mo	+56 mo
$1,800,000	90%	14 mo (1.2 yr)	+28 mo	+43 mo	+60 mo

Two things stand out.

The early-journey rows are gentler than the headline percentage suggests. At $200,000, a 40% crash destroys $80,000 and costs 17 months. That is because a small portfolio is contribution-driven: $48,000 a year of new money is enormous relative to the balance, so the hole refills fast.

The late rows are brutal, and in a way the "+43 months" understates. At 90% of target, the household was 14 months from freedom. A 30% crash makes it 57. The remaining journey does not lengthen by a third. It more than quadruples, because the loss is now large in absolute terms and contributions are small relative to it. There is nothing left to refill the hole except returns.

That asymmetry is the real content of the "sequence of returns" warning, and it is why the risk profile of a saver at 40% of target and a saver at 90% of target are barely the same problem.

The threat that does not look like a crash

Stoke's projection default is 7% real, which is a reasonable long-run historical figure. It is a much more aggressive figure for the next ten years specifically. Vanguard's Capital Markets Model, on its 30 June 2026 run, puts US equities at 4.2%–6.2% annualized nominal over ten years. Those are nominal figures. Against the 10-year breakeven inflation rate of roughly 2.25%, that implies something like 1.9% to 3.9% real.

So what happens to our $1,000,000 household if the crash never comes and the returns simply arrive smaller?

Scenario at $1,000,000	Months to FI	Versus baseline
7% real, no crash (baseline)	82 (6.8 yr)	0 mo
5% real, no crash	101 (8.4 yr)	+19 mo
4% real, no crash	114 (9.5 yr)	+32 mo
7% real, −30% crash today	117 (9.8 yr)	+35 mo
3% real, no crash	132 (11.0 yr)	+50 mo
2% real, no crash	156 (13.0 yr)	+74 mo

A 30% crash costs 35 months. A permanent shift to 3% real costs 50. The quiet scenario is worse than the loud one, and it is the scenario a major forecaster is currently projecting.

The cleanest way to say it: solving for the rate that reproduces the crash exactly, a 30% crash today is worth about the same as a permanent real return of 3.8% instead of 7%. Nearly the entire Vanguard-implied real band sits at or below that line. For this household, the return environment the market is presently pricing amounts to a 30% crash that has already happened and that will never show up on a statement.

But these are not two separate risks

This is where the analysis has to get more careful, because the obvious next move (add the crash to the low returns and despair) is wrong.

High valuations produce low forward returns. That relationship is the reason CAPE exists; Campbell and Shiller established it in 1988, and it is the finding underneath Shiller's Nobel. But a high CAPE resolves through one of two routes. Either prices fall and you take the pain quickly, or earnings grow into the prices and you take it slowly as years of thin returns. Both routes deliver a poor decade. They are substitutes, not additions.

Which means the useful question is not "will there be a crash?" It is: given a disappointing decade, which shape of disappointment do I want?

That question has an answer, and it is not the same answer for everyone.

Same ten-year return, different FI date

To isolate shape from magnitude, I built two paths with identical cumulative ten-year real returns:

Path A, crash first. A −30% shock now, then 6.74% real for ten years. Cumulative: 1.3439×.
Path B, slow grind. No shock, 3.0% real for ten years. Cumulative: 1.3439×.

Same decade, same total return, same ending index level. After year ten, both revert to 7% real. run_projection takes a single fixed rate, so this needed a return schedule. I extended the monthly recurrence and verified the extension reproduces production run_projection exactly on every constant-rate case across five balances and three rates before trusting it on the varying ones.

Portfolio	% of target	A: crash first	B: slow grind	Result
$200,000	10%	208 mo	221 mo	crash 13 mo sooner
$500,000	25%	169 mo	180 mo	crash 11 mo sooner
$1,000,000	50%	119 mo	127 mo	crash 8 mo sooner
$1,500,000	75%	79 mo	61 mo	grind 18 mo sooner
$1,800,000	90%	59 mo	24 mo	grind 35 mo sooner

The total return is held constant, so every month of difference in that table is produced by path alone. Partway down the column, the better choice reverses.

Early in the journey, you want the crash. Cheap prices are an asset when you have fifteen years of contributions still to deploy into them. This is sequence-of-returns risk running in reverse, the mirror image of the retiree's problem. Late in the journey, you want the grind, and you want it badly: at 90% of target the difference is nearly three years.

The crossover is five years, not a percentage

The obvious objection is that this crossover is an artifact of one household. So I solved for it across eight, varying spending from $50,000 to $120,000 and saving from $2,000 to $10,000 a month.

Annual spend	Saving/mo	Target	Crossover balance	% of target	Years to FI there
$80,000	$2,000	$2,000,000	$1,331,000	67%	5.0
$80,000	$4,000	$2,000,000	$1,201,000	60%	5.2
$80,000	$6,000	$2,000,000	$1,091,000	55%	5.2
$80,000	$8,000	$2,000,000	$988,000	49%	5.2
$50,000	$3,000	$1,250,000	$720,000	58%	5.2
$60,000	$2,500	$1,500,000	$935,000	62%	5.1
$120,000	$6,000	$3,000,000	$1,803,000	60%	5.2
$120,000	$10,000	$3,000,000	$1,582,000	53%	5.2

As a share of the target, the crossover wanders between 49% and 67%, too loose to be a rule. Measured in time, it barely moves at all: every household crosses over between 5.0 and 5.2 years from its FI date.

The usable rule:

Roughly five years out, the trade-off reverses. Further from your FI date than that, a crash is the better of the two ways an expensive market can disappoint you. Inside it, a crash is decisively the worse one.

It is not a coincidence. Five years is approximately how long your contributions need in order to exploit cheaper prices before you require the money. Past that point, you are functionally a retiree with a job.

What this changes, and what it does not

If you are more than five years out, the crash is not your problem, and bracing for one is not your best use of attention. Your exposure is the return assumption. Go look at the number in your own plan: if it says 7% real and you have never revisited it, you are carrying the difference between 82 months and 132 months as an unexamined default. Lowering it does not make you poorer. It makes your date honest, and it tells you now whether you need to save more or spend less rather than in 2033.

If you are inside five years, this is the zone where the mirror flips and you become vulnerable in the way retirees are. The levers are the ordinary ones and they are all better than prediction: a bond or cash allocation sized to the first years of withdrawals, flexible withdrawal guardrails instead of a fixed inflation-adjusted draw, and honest optionality about working a little longer. The table above already priced one version of that last lever: at 90% of target, a 30% crash costs 43 extra months of work. The question is whether you choose those months deliberately or have them chosen for you.

What this is not is a market-timing argument. Nothing here says a crash is coming, and nothing here recommends moving to cash. Sitting out is itself a bet on the shape of a decade, and this analysis is precisely about not having to make that bet. Bear markets are ordinary: 27 of them in the S&P 500 since 1928, averaging a 35% decline over about 289 days, as of Hartford's December 2024 tally. The point of the five-year rule is that you can act on your distance from the target, which you know today, instead of on the market's next move, which you do not.

Where this breaks

Four limitations, in rough order of how much they matter.

Contributions are assumed uninterrupted. This is the big one. Deep drawdowns correlate with recessions, and recessions cost people jobs. The whole reason the early-journey rows look mild is that $48,000 a year keeps arriving to buy cheap shares. A crash that also removes your income is a materially worse event than the one modelled here, and the model has nothing useful to say about it. Emergency reserves are the answer to that risk, not allocation.

Constant real returns are a modelling convenience. Real paths are lumpy. Holding the rate fixed is what makes the matched-path comparison clean, but no decade actually arrives as a flat line.

One portfolio, no frictions. A bond-heavy investor sees smaller shocks and lower returns, which moves both columns. Taxes, fees, and rebalancing are all ignored.

CAPE is a weak short-horizon signal and an imperfect cross-era one. It says close to nothing about next year, accounting standards have changed underneath it, and even Vanguard's own materials note that valuations predict poorly over shorter windows. It is evidence that forward returns are likely lower, not evidence about timing.

What to do this week

Find your return assumption and write it down. Not your target, not your balance. The growth rate your plan multiplies by. If you have never chosen it deliberately, you have inherited someone's default.
Compute your distance in years, not dollars. Percentage of target is the wrong ruler here; the five-year line is the one that changes what you should worry about.
Run the pessimistic version and look at the date. Not to act on it, but to know it.

The Can I FIRE Yet? calculator will do the third one without an account. It counts only FIRE-ready assets, and its Portfolio real return field is the input this whole article is about. Set it to 4% instead of 7% and watch the year move. To model the crash side, apply a haircut to the invested portfolio figure directly. It will not simulate a crash for you; you are applying the shock by hand, which is a fair description of the tool's current scope.

If a 30% haircut moves your date from 2033 to 2036, you have learned that you are on the safe side of the crossover and your attention belongs on the return assumption. If it moves it from 2027 to 2032, you have learned something considerably more urgent, and you have learned it while you still have five years of choices left.

This article is general information, not financial, tax, or investment advice. The households above are fictional and the projections are arithmetic under stated assumptions, not forecasts. Return estimates from any source, including the ones cited here, are projections and are frequently wrong. Consider professional advice for your situation.