Trusting the numbers

There are two ways to answer “will my retirement plan survive?”

Look backwards. Take the actual history of the U.S. market and replay it: how many real retirement start-dates would have survived? This is what the famous studies do — the Trinity Study, and Bengen’s original “4% rule.” It’s rigorous, and it’s the foundation of modern retirement planning. It also has a catch: the 20th-century U.S. market was unusually kind, and a backtest treats the luckiest run in developed-market history as the full range of possibility.

Look forwards. Simulate thousands of possible futures, including market sequences worse than any on record. This is what Futurez does, and it’s what the modern research firms do. It produces more cautious numbers, on purpose.

That distinction matters for reading everything below. If Futurez matched the backtests exactly, that would be evidence it was flattering you.

Morningstar publishes an annual study of the safe withdrawal rate — how much you can pull from your portfolio in year one, adjusted for inflation thereafter, and still have a 90% chance of lasting 30 years. It’s forward-looking, built on their own capital-market assumptions, and one of the most-watched numbers in the profession.

This is the single most meaningful validation we have: two forward-looking engines answering the same question, with no methodology difference between them.

We also checked Vanguard’s retirement calculator, using their own published example ($1.5M portfolio, $60,000/year, 30 years). Vanguard reports 91%; Futurez reports 89%. Vanguard’s engine draws on historical returns, which places it between the two camps — and a modest gap in the cautious direction is exactly what that predicts.

Across every published benchmark we tested, forward-looking and historical alike, Futurez lands at or below the published figure. Never above.

That’s the property that actually matters in a retirement plan. An engine that errs optimistic tells you to retire a year too early, or spend 10% too much, and you find out it was wrong at 82, when it’s too late to fix. We would rather be the tool that plays it a little safe.

Against the historical studies we run about five to six points below. Against Trinity’s 30-year 4% case, our results come out with a roughly 89% chance of success where the historical record says about 95%. We’re not going to explain that gap away. That gap is the whole point. It’s the difference between “this survived the 20th century” and “this survives most futures we can imagine, including bad ones.”

And we’ll show you our weakest result, too: against Trinity’s stock-heavy 75/25 case, we run about nine points below the published figure — our largest single gap in the suite. We’re showing it rather than hiding it, because publishing the weak spots is what makes the strong ones credible.

Matching a single number could be luck. The stronger test is whether an engine independently rediscovers the shape of what the research shows.

The Trinity Study found something that cuts against most people’s retirement instinct. The conventional move is to play it safe once you retire: shift out of stocks and into bonds so a market crash can’t sink you. But over a 30-year retirement with inflation-adjusted withdrawals, that “safe” choice is the riskiest one: a bond-heavy portfolio is the most likely to run dry, because its low returns can’t keep up with inflation while you draw an ever-rising income from it. Holding more stocks, up to a point, actually raises the odds your money lasts. Futurez was not designed specifically around this idea yet it produces the same curve: a steep climb up from a bond-only portfolio, then a plateau once you’re past roughly 60% stocks, with the cautious end of our curve sitting almost exactly on Trinity’s.

Everything above is about spending money down. Futurez also has to get building it up right.

Fidelity publishes a well-known guideline: save 15% of your salary and you should have roughly 1× your salary saved by 30, 3× by 40, 6× by 50, 8× by 60, and 10× by 67. Those milestones assume a target-date approach, a portfolio that de-risks toward retirement, shifting from stocks into bonds as you age, so that’s what we modeled, rather than a fixed allocation.

Do that, and Futurez tracks the line almost the whole way: 0.9× / 2.9× / 6.0× / 8.4× / 9.9× against Fidelity’s 1× / 3× / 6× / 8× / 10×.

Two honest caveats. Fidelity’s milestones are a deliberately conservative planning target — a floor you’re meant to clear, not a forecast. And we matched them with a moderate glide; a saver who holds more stock along the way would, in our model, finish above the milestones, because our returns are a touch more generous than Fidelity’s cautious assumption. Either way, a disciplined 15% saver reaches the mark.

This is the property that makes Futurez worth using to actually decide something, not just to read a number.

Take a retiree drawing $40,000 a year with an 89% chance of lasting 30 years, and make one ordinary change: trim spending by $200 a month, roughly one dinner out a week, about a 6% cut. The chance of success rises to 92%.

The figure that matters there isn’t the 89 or the 92 — it’s the three-point gain. And you can trust it, because both versions of the plan are scored against the same 1,000 market futures. Nothing is reshuffled between runs; the only thing that changed is the spending. Run either version again and you get the identical answer.

Contrast that with a simulator that draws fresh random markets on every run. There, the same good decision can make the number fall, not because it was a bad decision, but because a new roll of the dice happened to land worse, and you’d have no way to tell a real improvement from a reshuffle. Every success rate, ours included, carries a couple of points of sampling wobble. What Futurez does is keep that wobble out of the comparison, so a move of a few points is a signal you can act on.

So read the number the way it’s meant to be read: don’t agonize over 85% versus 87%. Ask what moves it, and trust what the tool tells you about the size and direction of that move. That’s the question a plan is actually built to answer.

The benchmarks above test the engine’s output. Underneath them is a market model calibrated against roughly a century of U.S. market history, real returns, real crashes, real inflation, with volatility, fat tails, clustered downturns, and stock/bond behavior all tuned to match the record. On top of that, we deliberately mark near-term stock returns down toward the cautious forecasts that major asset managers publish today, because current stock valuations are historically high. In plain terms: the model’s behavior is grounded in history, and its near-term return outlook leans conservative on purpose.

• These are probabilities, not promises. No simulation knows the future. A 90% plan fails one time in ten, and that’s the point of the number.
• Reproducible isn’t the same as certain. Your plan always scores the same, and that’s a real feature, but it doesn’t make the number exact. It’s still one model’s estimate, resting on assumptions about future returns, and those assumptions do the heavy lifting. That’s why we report whole percentages and never decimals: a decimal would be stable, but it wouldn’t mean anything.
• We didn’t match every benchmark, and we showed you the misses. The stock-heavy Trinity case is our largest gap, at about nine points below.
• Benchmark comparisons are approximate. Our portfolio settings don’t map perfectly onto the exact stock/bond mixes each study used, so a point or two of every difference above is that mismatch rather than the engine.
• This isn’t financial advice, and past performance, including 100 years of it, can’t guarantee future returns.

Tested against the retirement industry’s published research, Futurez matches the leading forward-looking study almost exactly, reproduces the relationships the historical studies found, tracks the industry’s savings milestones, and on every single benchmark never once claimed your plan was safer than the research says. That’s what we mean when we say you can trust the number.