# All-weather portfolios: Browne, All Weather, Golden Butterfly, Dragon Markets have seasons, and no asset wins in all of them ([[market-regimes]]). One conclusion follows. Instead of betting on the likely season, you hold at least one winner per season at all times and let rebalancing harvest the rotations. That simple idea produced a line of famous portfolios: Harry Browne's Permanent Portfolio (1981), Ray Dalio's All Weather (Bridgewater), the Golden Butterfly from the Portfolio Charts community, then Artemis Capital's Dragon. They are all cousins. They all rest on the same growth × inflation grid. And they all sell the same product: a **tighter** distribution of returns, with no fat left tail (the flank of the distribution where the extreme losses live, [[fat-tails]]), paid for with a lower expected return than a portfolio dominated by stocks. That is exactly the trade the math of withdrawal rewards ([[arithmetic-vs-geometric-returns]], [[sequence-of-returns]]). They are mediocre vehicles for accumulation and excellent candidates for **decumulation**. Their historical withdrawal rates hold up next to those of portfolios that look far more impressive on paper. This article walks the family portfolio by portfolio, with exact compositions, the logic of each and the real track records. It gives the numbers that matter to a retiree: the SWR, the real drawdowns, the worst decades. It weighs the serious criticisms: the reliance on gold, the long duration, and the gap to the index that is so hard to hold. And it ends on the dosed version most plans should prefer to all-or-nothing, and on how to put that version to the test. ::: cle Why a retiree cares For an **accumulator**, an all-weather portfolio is a brake. It returns 4 to 6% real against 6 to 7% for an equity portfolio, which compounds into decades of lost ground. For a **retiree**, the scale changes. What governs the withdrawal rate is not the average but the **worst path** ([[the-trinity-study]]). And all-weather portfolios have remarkably gentle worst paths. Their maximum real drawdowns run from −20 to −30%, against −40 to −50% for an 80/20, and the history available shows no deep lost decade. Since 1972, their SAFEMAX has matched or beaten that of portfolios far more profitable on average. You are not buying their mean. You are buying their floor. ::: ::: figure tous-temps-saisons The four compositions detailed below, weights unchanged, colored not by asset but by the season each sleeve is there to cover. Assigning a sleeve to a season is an editorial reading of the growth × inflation grid, and each sleeve counts for one season only. ::: ## Browne's Permanent Portfolio: the archetype Harry Browne, investment adviser, libertarian essayist and twice a US presidential candidate, published this portfolio in 1981, then popularized it in *Fail-Safe Investing* (1999). It fits in four quarters: **25% stocks, 25% long government bonds, 25% gold, 25% cash (money market and T-bills)**, rebalanced as soon as a sleeve drifts outside the 15 to 35% band. The match with the regime grid is explicit, and it is the founding one. Stocks cover prosperity, long bonds deflation, gold inflation and crises of confidence, and cash covers the recession while funding the purchase of everything else. Browne does not claim to predict. He claims that prediction **is not possible**, and draws from that the allocation of maximum ignorance, weighted equally across the four answers. The real record rests on US data from 1972 to 2024, since gold only floated after 1971. It comes to a little under 4.5% real a year for a volatility of 7 to 8%. The **worst** real year is around −17%, in 2022, under the one regime that hits three sleeves at once: inflation plus rising rates. The maximum real drawdown reaches about −22%, spread over the 2021-2022 inflation surge. That is the hardest episode, the one where cash and bonds were losing in real terms despite gold. The early 1980s produced one almost as deep, when gold lost half its value after its January 1980 peak. In 1973-74, 2000-02, 2008 (−2% nominal!) and 2020, the portfolio came away with scratches where equity portfolios were bleeding. The historical SWR comes out around 4.5 to 5% over 30 years, above the 60/40. The usual warning still holds: this 50-year sample contains a single great inflationary regime and forty years of a bond bull market ([[simulator-traps]]). ## The family: All Weather, Golden Butterfly, Dragon **Dalio's All Weather** (Bridgewater) was formalized in the 1990s, then packaged for the public by Tony Robbins as roughly 30% stocks, 40% long bonds, 15% intermediate bonds, 7.5% gold and 7.5% commodities. It refines Browne with **risk weighting** (risk parity). The idea is to balance the risk contributions of each season rather than the capital. Bonds, less volatile, get more capital. The institutional version uses leverage to bring the expected return back up to equity level. The retail version does without, hence a modest expectation, about 5% real, for a volatility of 8%. Its own weakness broke out in 2022: the massive overweight in duration. When inflation hits, the 55% of long and intermediate bonds take the shock together, −20% nominal, the worst year in its history. It is the reminder that "weighted by **past** risk" stops meaning "all-weather" once correlations flip ([[market-regimes]], [[bonds-in-retirement]]). **The Golden Butterfly** (Portfolio Charts, Tyler) adds a fifth wing to the Permanent: **20% broad stocks, 20% small-cap value, 20% long bonds, 20% short bonds, 20% gold**. The idea is to tilt the portfolio toward prosperity, the most frequent regime, through the small-cap value premium ([[factors-in-retirement]]), without giving up the other three seasons. Over 1972-2024 it returns about 5.8% real for a volatility of 8%, with a maximum real drawdown of about −22%. It is the best return-to-worst-path pair in the family on the sample available, and it regularly posts the highest SWR on all of Portfolio Charts, around 5.5 to 6% over 30 years in the United States. Read it with the suspicion owed to backward-looking optimization, because the choice of small-cap value is guided by the sample. **Artemis's Dragon** (Chris Cole, "The Allegory of the Hawk and Serpent", 2020) is the extreme version, and the most interesting one intellectually. It runs roughly 24% stocks, 18% long bonds, 19% gold, 18% commodities and trend, and 21% long volatility, the option strategies that **make money** in crashes. Cole rebuilds a century, 1928 to 2019, and shows that such a mix gets through **every** great destruction, the deflation of the 1930s, the inflation of the 1970s and the crashes, where the 60/40 alternates glory and ruin by the decade. One practical limit rules it out for an individual. There is no cheap, accessible fund for the long volatility sleeve, and what exists degrades into expensive products; [[long-volatility]] walks that wall end to end. The buildable version replaces long volatility with managed futures ([[managed-futures]]), its workable cousin. You give up the convexity of an instant crash and keep the cover for the long regimes. ::: science What the numbers really say, and their limits Over 1972-2024 (US series, an independent reconstruction cross-checked against Portfolio Charts), in real terms: | Portfolio | Return | Max drawdown | 30-year SWR | |---|---|---|---| | 60/40 | 5.6% | −38% | ~4% | | Permanent | 4.4% | −22% | ~4.5-5% | | Golden Butterfly | 5.8% | −22% | ~5.5% | | All Weather | 5.1% | −29% (2022 included) | ~4.5% | The robust lesson comes in three findings. At equal return, the worst path is almost twice as shallow: the Golden Butterfly returns like a 70/30 while diving only 22%, where that 70/30 dives 42%. No mix of stocks and bonds, however cautious, ever did better than −36%. And it is the worst path that sets the withdrawal rate. The limits are just as robust. We are working on **one** 50-year sample, dominated by a secular bond bull market and by a gold price that started from administered levels in 1972. Its 1970s decade at +1,300% nominal was a liberation event, not a permanent property of the asset ([[gold-in-retirement]]). Outside the United States, the replications give the same **shape**, short tails, at more modest levels. Caution is in order: keep the structure, mark the levels down, test against the world century. One practical reservation on top of that. The long multi-country panels are built on domestic stocks and bonds, and they barely let you simulate anything other than a 60/40 ([[anarkulova-cederburg]]). They test the shape of a withdrawal plan, never one particular allocation recipe ([[historical-vs-parametric]]). ::: ::: figure tous-temps-echange Each portfolio becomes a point: annualized real return on the x axis, worst real path on the y axis. The all-weather family sits in a corner no stock-and-bond mix ever reaches, and the Golden Butterfly returns like a 70/30 for a dive almost twice as shallow (US, 1972-2024, real, annual rebalancing). Its small value leg pays 1.0 point of costs a year, which Kenneth French's academic portfolio does not pay. ::: ## The honest criticisms, and who they are aimed at **"Half the portfolio produces nothing" (gold and cash in Browne).** True in terms of cash flow, no coupon and no earnings, and false in terms of the portfolio. Those sleeves buy cover for the regimes where productive assets die, and their "return" is harvested at rebalancing: selling the gold of 1979 to buy the stocks of 1980, selling the bonds of 2008 to buy back the stocks of 2009. The criticism still puts a lid on the expected return. An all-weather portfolio will **never** compound like an equity portfolio, and the young accumulator who adopts one is paying for insurance he does not need ([[the-three-phases]]). **The reliance on gold.** That is 20 to 25% of the portfolio in an asset with no cash flow and no falsifiable valuation, whose modern sample is short. This is the family's real concentration of **epistemic** risk ([[gold-in-retirement]]). Modern variants cut it by splitting the inflation sleeve between gold, inflation-linked bonds (linkers, [[inflation-linked-bonds]]) and commodities and trend. Three imperfect hedges beat one. **The long duration.** Browne and Dalio want **long** bonds, for maximum deflationary punch, and 2022 showed the bill in the opposite regime. The modern compromise is simple: mostly intermediate duration, plus a long slice owned openly as "deflation insurance", the way gold plays "inflation insurance". **The long sample, the most recent criticism.** It comes from ERN's Part 64 (2026). Jeske replays All Weather, the Golden Butterfly and a cousin, the Golden Ratio, on monthly US data reaching back well before 1972. He first strips out the small-cap value premium, to judge the construction alone. Over 30 years the three portfolios then hold between 3.35% and 3.51%, against 3.83% for a plain 75/25. After 1926, All Weather and the Golden Ratio edge back in front of it. Over 50 years with a quarter of the capital left to heirs, the gap widens: 2.45% to 2.71% against 3.24%. Jeske puts it down to too few stocks: over a horizon that long, he argues, the shortfall costs more than the short tails bring in. With the historical small-cap value premium, the two "Golden" portfolios are competitive again after 1926. But on his reading that gain comes from small-cap value, not from balancing the seasons, and a 75/25 given the same dose gets more out of it ([[factors-in-retirement]]). Two reservations temper this replay. Before 1971 the price of gold was officially fixed, so the asset replayed is not the one you can buy today. And the small-cap value premium is either left whole or set to zero, when the answer probably sits in between. So the two windows correct each other: the one that opens in 1972 flatters these portfolios, and ERN's works against them ([[the-ern-series]]). The criticism is also aimed at the full recipes, held for a lifetime. Jeske himself would rather add one building block to a 75/25, gold for instance, which is the dosed use described below. **The gap to the index, the most serious criticism.** By construction, an all-weather portfolio makes +6% in the year the world makes +25%. And the ordeal is counted in years rather than in points: over the fifty-three years reconstructed here, the Golden Butterfly spent thirty-four of them behind stocks, including one run of nine in a row. Holding that "underperformance" **for years**, while reading everywhere that stocks are flying, is a documented behavioral ordeal. Most all-weather portfolios are abandoned at the top of an equity bull market, which is to say at the worst possible moment ([[the-psychology-of-spending]]). All-weather asks the same thing a glide path asks ([[glidepaths]]): a written commitment, for written reasons, reread when the itch comes ([[building-your-plan]]). ::: figure tous-temps-ecart The annual real return gap between the Golden Butterfly and 100% stocks, year by year (US, 1972-2024, real, annual rebalancing, the small value leg net of 1.0 point of costs a year). The gray bands mark the runs of at least four consecutive years behind. ::: ## The recommended use: a dose of all-weather, sized to the plan The choice is not binary. Modern design treats all-weather as a **dial**. Between a growth portfolio (global stocks and bonds, the core of [[stock-bond-allocation]]) and a full all-weather portfolio, every weighting exists, and the right dose depends on the phase and on the floor. The practical markers are clear. In **accumulation**, the dose stays small or zero. In the **transition** and through the first decade, the fragile window, it rises. Call it a glide path of structure, and it completes or replaces the bond tent ([[glidepaths]], [[the-three-phases]]). In the **covered phase**, set it by temperament. In practice, many robust plans converge on a core of 60 to 70% global growth plus 30 to 40% in a regime sleeve (gold, linkers, long duration and possibly trend). This half all-weather captures most of the tail shortening for 0.3 to 0.6 points of expected return. ::: figure tous-temps-curseur The dial dose by dose, on the 70/30 core of the worked example below: before the plateau, each slice of sleeve buys a lot of worst path for almost nothing; inside it, the buying stops while the return keeps being paid. The sleeve here is half gold, half long duration, the example's linkers having no real-dollar series available. US, 1972-2024, in real terms, annual rebalancing: this is a shape, not a universal level. ::: ::: astuce Test the variant before you adopt it An all-weather portfolio is judged A/B, never in the absolute. Build the variant from the blocks you can actually buy (global stocks, long and intermediate bonds, gold, linkers, managed futures; [[building-it-with-us-etfs]] goes through the shopping list) and run it against your current portfolio, same withdrawal rule, same horizon. The readings that decide are the tails, which is where these portfolios do their work: a sequence stress, a lost decade, and a replay of the 1929 and 1966 vintages. Two less obvious effects deserve a look. The cash buffer you need **shrinks** when the tails are short, since the insurance already sits inside the portfolio. And a flexible rule fires fewer cuts, so the standard of living delivered is steadier. Expect the typical pattern: central-case failure probability equal or a little better, clearly better tails, lower median ending wealth. That is the contract, printed on the label. For this family, portfoliocharts.com already publishes the SWRs and drawdowns country by country, enough to frame the comparison before you simulate anything. ::: ::: exemple A half-Browne put to the test The plan: $1.5M, $51,000 a year (3.4%) with a Vanguard corridor, over 45 years, a pension arriving in year 15. Variant A: a classic 70/30 with intermediate bonds. Variant B: 60% global stocks, 15% intermediate bonds, 10% long bonds, 10% gold, 5% linkers. The verdicts come in. Central-case failure probability 4.2% for A against 3.6% for B. Sequence stress 7.5% against 5.4%. Lost decade, B survives outright, and the maximum real drawdown of the median paths goes from −38% to −27%. Median wealth at 45 years, $3.1M for A against $2.6M for B. The couple picks B for the fifteen years of uncovered phase, the bridge years before the pensions start, with one written clause: revisit the gold dose once the pensions are claimed. No dogma and no all-or-nothing, just insurance that is sized, dated and reversible. ::: ## The essentials - The family (Browne's 4 × 25, the risk-parity All Weather, the Golden Butterfly, the Dragon) applies the regime grid: one winner per season, rebalanced on a schedule. It sells short tails (real drawdowns of −20 to −30%) for a modest expected return (4.4 to 5.8% real). - For a retiree the trade is the right one, because the withdrawal rate depends on the worst path, not on the average. All-weather SWRs match or beat those of far more impressive portfolios, on a 50-year sample that deserves a markdown (bond bull market, post-1971 gold). Replayed on the long sample by ERN (Part 64), the full recipes do worse than a 75/25, above all over 50 years, with gold at an officially fixed price before 1971. The answer sits between the two windows. - The real criticisms are known: the reliance on gold (split the inflation sleeve), the long duration (2022) and above all the gap to the index during the bull years. The commitment gets written down before it gets lived. - The modern use is a **dial**, not a dogma. A growth core plus 30 to 40% in a regime sleeve, the half all-weather, captures most of the short tails for 0.3 to 0.6 points of expected return, with the largest dose during the fragile window. - Run the A/B under the hard models, sequence stress and lost decade, and over the 1929 and 1966 vintages. That is where these portfolios do their work. Accept the lower median wealth in exchange: it is the advertised price of the contract. --- ## Going further - Harry Browne, *Fail-Safe Investing* (1999); Rowland & Lawson, *The Permanent Portfolio* (2012): the archetype and its manual. - Bridgewater, "The All Weather Story" (bridgewater.com); Chris Cole (Artemis), "The Allegory of the Hawk and Serpent" (2020): the institutional versions. - [portfoliocharts.com](https://portfoliocharts.com): the histories, SWRs and drawdowns of the whole family, country by country, the free reference tool. - Early Retirement Now, Part 64 (risk parity in retirement): the criticism on long data ([[the-ern-series]]). - In this book: [[market-regimes]] (the founding grid), [[gold-in-retirement]], [[inflation-linked-bonds]], [[managed-futures]] (the building blocks one by one), [[defensive-assets]] (the map of the roles), [[long-volatility]] (the Dragon's volatility sleeve in detail), [[why-diversification-works]] (the rebalancing premium that keeps these recipes alive).