The numbers, with their methodology attached.

The Distillate: four modules, one discipline.

No patterns, no predictions, no indicators. The engine acts only on the one thing our research showed to be forecastable — volatility — and everything else exists to keep it honest.

Data → intelligence → decision.

The numbers, with their methodology attached.

How these figures are computed. The historical engine figures come from a full-history test that decides from the previous close, derives its position from what the account holds each session, buys only with settled cash under the executor’s own budget, and keeps its 2% sizing reserve, the way the live executor does. Whole shares, $1 a leg, dividends included, modelled next-close fills rather than the 7:45 AM Pacific execution price. Idle and parked cash earn the historical three-month Treasury yield, credited on the previous close’s settled balance — a rate assumption, not a model of any cash fund’s shares, distributions or fees. A worst fall is what this test produced, not a limit on what an account could lose; it moves about two points on the cash assumption alone, because a different cash balance buys a different whole number of shares. The full-history test and its frozen inputs are in the repository; the ledger entry is a test whose rules were written down in advance. The same figures are drawn as engine cards.

The deep tables below measure every engine. Every claim we publish falls into one of four evidence grades, and every grave in the graveyard is stamped with the grade of its own autopsy. Most products only ever show you C and call it S.

swipe for all columns →

What the safe option costs.

Buy & hold is the academic comparison. The one most people actually face is leaving the money in savings. Here it is — measured in what the money will buy, not what the statement says.

Read the drawdowns before the returns. Since 2009 a NASDAQ rule with no exit and QQQ buy & hold finish almost together — in this sample, that rule offered little separation from the index on return. The exit engines separate from both — SELECT 18.5% and ULTRA 22.2% a year since 2009 — at a third to half the drawdown. The separation is in the other column. Over the full cycle, buy & hold’s worst real drawdown is −84% — $5,000 of purchasing power becoming about $800 before recovering — and the historical record is full of accounts that sold there. ULTRA’s worst was −30%, SELECT’s −24%, STEADY’s −19%. Buy-and-hold returns exist on paper; surviving the path is the product.

Inflation averaged 2.60% a year across both windows, so a savings account has a negative real return in both: $5,000 left in cash since 2009 buys about $4,271 of goods today. Measured in purchasing power rather than dollars, cash’s worst drawdown since 1999 is −23.9%, it spent 94% of all days below a previous high, and its longest stretch underwater was 17.6 years — longer than any equity drawdown in this sample. “Cash has no drawdown” is an artefact of measuring in the unit that is shrinking. We use the 3-month T-bill because it moves with rates the way savings rates do; note it understates a good savings account in 2009–2015, when top accounts paid near 1% and bills paid almost nothing.

What compounding looks like.

Illustrative compounding — not an account-size backtest. These tables apply the same historical rate at every starting balance. Actual results also depend on whole-share rounding, costs and cash settlement.

Compounded at the full-cycle 2000–2026 rates — the same window as the chart below, dot-com crash included. The buy & hold column is the honest yardstick — the gap between it and the three engines is what the discipline is worth, and it is not a promise: history is one sample, and every mode spends years underwater along the way. Watch the live line rather than trusting a table.

The same $5,000, four disciplines, through the dot-com crash, 2008, COVID and 2022. The lower panel is the honest half: how far below its own high-water mark each one sat, every day. STEADY never went deeper than -18%; buy & hold spent years past -54%. ULTRA's -26% is the price of its extra compounding — it earns the most of the three and it hurts the most. Log scale — equal vertical distance is equal percentage move, which is the only honest way to draw 26 years. Notice where the lines separate: not in the bull runs, but in the craters.

The dashed line is the one to respect. Buying and holding the S&P 500 is the gold standard of investing — roughly 9–15% a year depending on the era, for zero effort, and most professionals never beat it. It is the default your money already deserves, and the only honest reason to run anything more complicated is what this chart shows: beat it, or take meaningfully less pain earning it. That dashed line is the bar we measure ourselves against — never the zero line.

What a monthly habit builds.

Same $3,000 start, same full-cycle rates as the projection table above. “You put in” counts the start plus every deposit — $50 a month for ten years is $3,000 + $50×120 = $9,000 of your own money; the strategy’s work is the distance between that column and its own. Deposits buy toward target on the next run, so every contribution is also a free rebalance.

After five years · starting from $3,000

swipe for all columns →

After ten years · starting from $3,000

swipe for all columns →

The “none” rows are the strategy alone — identical to the projection table above, same math. At this account size the habit is worth more than the strategy: going from nothing to $50 a month roughly doubles the ten-year outcome. Held for twenty years, $200 a month at the same rates builds roughly $197,000 on STEADY, $263,000 on SELECT and $364,000 on ULTRA from $51,000 contributed. Same caveat as everything on this page: compounded history, not a promise.

Ready to decide? The engines and the price are on the front page. Not convinced? Read what died first.

“Under observation” does not mean trading. “Applied improvement” describes completed research or implementation work, not approval to sell an expansion. R-numbers are reference IDs.

Loading the research ledger…