Lean Sixty

A 60/40 held in two-thirds of the money, so the freed third can hold diversifiers for when stocks crash. Long Treasuries are held for crashes that bring falling rates, gold for the ones driven by inflation or a weak dollar. In the record they cushioned 2000 and 2008 but not 2022, when stocks and bonds fell together. Three ETFs, fixed weights, rebalanced by dividends and a monthly buy-the-dip check. Designed for a taxable account. Published with the full test record, including the variants that failed.

It is an alternative to the popular 60/40 and three-fund portfolios (total U.S. stocks, total international stocks, total bonds), designed for more after-tax return with shallower declines. Unlike a three-fund, it holds no international stocks.

Status: backtest only, not yet traded. The live record begins on launch. Every figure on this page is a historical simulation, not a prediction, and is written from the results files by script (render_lean_sixty.py). Record 20e4864 · data snapshot 2026-07-09-prototype · recorded 2026-08-21.

What it holds

The rule, in two parts. Dividends (and, in a live account, new contributions) buy the sleeves that are below target, in proportion to their shortfall. Once a month, if the stock sleeve (NTSX) has fallen more than 5% of the portfolio below its 67% target, or more than 25% of its own weight, it is bought back to target: first with any cash, then by trimming gold and long Treasuries down toward their targets, never below them. Those trims are the only sales the portfolio ever makes. Sleeves that drift above target are otherwise left alone, so realized gains are rare: the record shows 3.3% annual turnover over 26.0 years, with 0.10% per trade charged as cost.

Lean Sixty Reserve is the same portfolio holding a 10-point Treasury-bill reserve with written firing rules: the reserve is spent into the portfolio in three steps as the portfolio falls 15%, 25% and 35% from its high, and refilled only from dividends afterward.

Why “lean”, and the leverage arithmetic

A plain 60/40 uses every dollar to hold 60 cents of stocks and 40 cents of bonds. Lean Sixty gets the same stock-and-bond shape from 67% of the money: that sleeve alone delivers about 60% in stocks and 40% in Treasury exposure. The remaining 33% is free to hold gold and long Treasuries. Same core, fewer dollars: that is what “lean” means here.

How that 33% was split. By inverse volatility, on long-run volatility figures written down before any test was run: gold at 16%, long Treasuries at 12%. The more volatile sleeve gets fewer dollars, so neither dominates the bucket’s behaviour — gold’s share is (1/16) ÷ (1/16 + 1/12) = 42.9%, which is 14.1 points of the 33%, leaving 18.9. Rounded to whole percentages for a brokerage pie: 14% gold, 19% long Treasuries.

Two things follow from that, and they are worth being plain about. The rule balances risk, and makes no forecast about returns — so it should be expected to land near the mix that minimises declines, and to have no opinion about the mix that maximises return. That is what the record shows. And because the weights were fixed by a stated rule in advance rather than chosen after seeing results, their backtested figures are an estimate of what that mix does, rather than the best of many mixes that were tried. Mapping the alternatives afterward (see the ledger below) found mixes that scored better over this window; they are not held, because a mix picked for scoring well on a window has no evidence behind it beyond that window — and when those same mixes were tested on earlier years, the ranking reversed.

The Treasury exposure inside NTSX is held through futures, so the portfolio’s gross notional exposure is about 134% of its value. Say it plainly: this is leverage, contained inside one fund. Three things follow.

• It is not a margin loan. There is no personal borrowing and no margin call; the fund cannot lose more than its value, and the portfolio holds it at a fixed weight.
• The cost is the financing rate embedded in the futures, roughly a T-bill yield on the extra 60 cents, plus the fund’s expense ratio.
• The risk is a year when stocks and bonds fall together. That happened in 2022, and it is this portfolio’s worst case. The numbers are below.

The record

Window 2000-06-30 to 2026-06-30 (26.0 years). “60/40” is a buy-and-hold position in Vanguard’s Balanced Index Fund (VBINX: 60% U.S. stocks, 40% U.S. bonds), taxed the same way. Wealth is after-tax liquidation value per $100,000: every lot is tracked, dividends are taxed by their character each year, and the portfolio is sold and taxed at the end. The headline bracket pair is 35% ordinary / 20% long-term gains; the second pair is 24% / 15%.

Against a three-fund portfolio. The three-fund here is iShares Core Growth Allocation (AOR): a 60/40 of U.S., international and bond index funds in one ticker, held the same tax-efficient way as the 60/40. It is the lab’s standing benchmark row, computed before this page used it; this comparison was added after the record, changes no verdict and selected nothing. Read three things first:

• Over this window international stocks lagged U.S. stocks, so the three-fund trails the U.S.-only 60/40 by 0.50pp/yr before tax. Part of Lean Sixty’s wider margin here is that lag, and Lean Sixty holds no international stocks at all. A period when international stocks lead would narrow it.
• AOR’s own returns are used from 2009-06-30. Before that the row is a fitted blend of index sleeves (U.S. 39%, international 18%, emerging 6%, bonds 37%; correlation 0.996 with the fund) that charges no expense ratio, which flatters the three-fund slightly.
• Only full-window figures exist for this row. The crisis-by-crisis comparison, 2022 included, is against the 60/40 below.

Where the edge comes from. Against the same three sleeves without the capital-efficient core, the core adds $140,109 per $100,000; against the same portfolio rebalanced with contributions only, the monthly buy-the-dip check adds $48,705 on top.

Rolling ten-year windows (193 of them, monthly starts): Lean Sixty finishes with more after-tax wealth in 100% and with a shallower worst decline in 69.9%. The wealth claim is the robust one. The drawdown claim depends on which crisis the window contains.

Sub-periods (after-tax CAGR at 35/20; worst decline pre-tax):

The three crises, measured from each prior peak (the sub-period table above measures declines only within each period, so 2022 reads deeper here):

What the record does not show, stated first

• 2022 is the failure mode. When stocks and bonds fell together, Lean Sixty fell -29.2% against the 60/40’s -21.6%. Every rolling window in which it had the deeper decline is one that contains 2022 without 2008. The shallower full-window decline is a 2008 story.
• Most of the history is a synthetic fund. The record uses the real fund from 2018-08-03; before that the sleeve is a validated synthetic (90% stocks plus 60% Treasury futures net of financing and fees; correlation 0.962 with the real fund over their overlap, and the real fund ran 0.60pp/yr above it, so the backfill is conservative), but the independent audit found 72.2% of the wealth edge accrues in the synthetic era. On the real fund’s own years the wealth edge is 1.91pp/yr, while the drawdown edge reverses: −28.9% against the 60/40’s −22.8% (audit figures, quoted).
• The long Treasury sleeve’s crisis value is a pre-2018 measurement. Varying that sleeve across 1,221 weight combinations shows it doing real work over the full record: dropping it entirely deepens the worst decline by 9.1pp, which it pays for with 0.67pp/yr of return. On the real fund’s own years that reverses — the worst decline barely moves however much is held (0.4pp across the whole range), while holding the sleeve costs 1.55pp/yr and 0.21 of return-per-decline. The sleeve is held because two full crises say it earns its place; one recent window says it did not. Both are in the record.
• The weights are not a fitted peak, but they are not arbitrary either. Across those 1,221 combinations the published mix ranks near the middle on every measure (41% on wealth, 45% on return-per-decline, 61% on the worst decline), so the record is not the product of a lucky choice — the weights were set from long-run volatility before any of it was run, and were never tuned. They do matter, though: moving gold or long Treasuries by four points either way spans 0.21 of return-per-decline around a median of 0.99. Treat the headline figures as the middle of that range, not as a precise forecast.
• The drawdown margin passed thin and depends on the basis. Scored on pre-tax value, as pre-registered, the decline was +5.5pp shallower against a 5pp bar. Scored on after-tax paths it is +4.745pp, a miss by a quarter point (audit figure, quoted). The Sortino and UPI multiples also cleared their 1.25× bar narrowly.
• The 1993 extension. On a window starting 1993-06-30, the wealth edge is +1.61pp/yr and the decline edge +5.5pp, so both survive out of era. The Sortino and UPI multiples do not (1.12× and 1.24×); they partly measure how poor 2000–2026 was for a 60/40. Lean Sixty trailed the 60/40 on the 1993–2000 leg by 0.59pp/yr.
• The window flatters diversifiers. It starts at the dot-com top and near gold’s secular low.
• One fund, one issuer. The core sleeve is a single WisdomTree product. That is an operational risk no backtest measures.
• Execution. The record assumes month-end trades. A cell that trades on the next day’s prices ends at $733,183, so timing within a day or two does not drive the result. A cell taxing every NTSX distribution at ordinary rates ends at $707,809.

Lean Sixty Reserve

The reserve version ends at $717,425 per $100,000 at 35/20, $3,846 less than Lean Sixty over the full window (−0.02pp/yr), for a worst decline of -28.5%, an Ulcer Performance Index of 1.07 against 0.99, and a shallower trough in every crisis. It paid tax on $8,598 of reserve interest along the way.

What the reserve’s own path shows: it fires about once a decade. 2008 drained it to 1.8% of the portfolio, dividends took most of a decade to refill it, and 2022 met a partly refilled reserve and drained it again; it stood at 4.0% on 2026-06-30. Because the rules are checked monthly, a fast intramonth fall can pass unseen; the dot-com decline never tripped the first threshold on a month-end.

The ledger: how it was tested, and what failed

The program pre-registered each round’s hypothesis, cells and acceptance bars before running it, and refused near-misses rather than tuning toward the bar.

The round log, pre-registrations and result files are kept in the lab’s research repository, which is not public. The figures file names the result file behind every number on this page.

Implementations

Tested: the three ETFs above. That is the record on this page.

Untested: a futures replication. A trader can build the same exposures directly: the stock leg in an index fund or S&P 500 futures, the 60-point Treasury leg as a ladder of Treasury futures rolled quarterly, and gold in an ETF or gold futures. This removes the single-issuer dependency and the fund’s expense ratio, at the cost of roll spreads, commissions and the financing rate implied in the futures. It also changes the tax character: futures are marked to market every year and taxed at a blended rate with no deferral, while the record above earned its after-tax edge under ETF treatment, where gains defer until sale. For the Treasury leg that is worse; for gold, taxed as a collectible in an ETF, it may be better. The pre-2018 synthetic in the record is itself a futures construction and tracked the real fund closely, but that supports the pre-tax tracking only. No after-tax record exists for this implementation, and none is claimed. It would be a new pre-registered round. In a tax-deferred account the tax difference disappears and the futures version is the cleaner one to test first.

Who this is for

Someone who would hold a 60/40 or a three-fund portfolio in a taxable account for a decade or more, who wants a gold and long-Treasury sleeve without giving up the stock-and-bond core, and who has looked at the 2022 row above and can hold through a year like it. It is a whole-portfolio alternative to a 60/40 or a three-fund, not a satellite.

It is not for anyone who wants international stocks in the portfolio, who wants a plan without a futures-based fund in it, or who would sell in a drawdown deeper than a plain 60/40’s.

Reproduce it

Weights and rule as above; data snapshot 2026-07-09-prototype; results recorded at commit 20e4864 of the lab’s research repository (not public); the figures file lists the source of every figure. Metric definitions are on the methodology page.

Run this mix yourself in the Backtest tool (calendar rebalancing, not this page’s rule), or replay a retirement plan holding it over every historical start month in the retirement stress test.

Hypothetical performance. Historical simulation from ETF and synthetic price data, net of modeled costs and taxes; not a prediction, not advice. See disclosures and methodology.

Rendered from record 20e4864 on 2026-09-29; 117 figures, each with a source file listed in the figures file.