Permanent Portfolio vs. Golden Butterfly vs. Cockroach: Who Each One Is For

Three portfolios come up whenever investors ask how to stay invested without guessing the next recession or inflation scare. Harry Browne’s Permanent Portfolio splits money four ways. The Golden Butterfly, designed by Tyler at Portfolio Charts, adds more stocks and a small-cap value tilt. Mutiny Fund’s Cockroach adds trend following, long volatility, and leverage. All three promise a smoother ride than a stock-heavy portfolio, and all three ask you to give up some expected growth for it.

We rebuilt the Permanent Portfolio and the Golden Butterfly from primary data for 1972 through 2025, split the record before and after each was published, removed one sleeve at a time to see what each contributed, and estimated what taxes cost in a brokerage account. The Cockroach cannot be rebuilt that way, because its manager does not publish its weights or its returns, so it gets a different kind of review below.

What each portfolio holds

The Permanent Portfolio holds 25% each in US stocks, long-term Treasuries, cash (T-bills), and gold.¹ Browne’s logic was that each asset does well in one economic environment: stocks in prosperity, long bonds in deflation, gold in inflation, and cash in a tight-money recession. Because nobody can predict which comes next, you hold all four and rebalance.²

The Golden Butterfly holds 20% each in a US total market fund, US small-cap value, long-term Treasuries, short-term Treasuries, and gold.³ It moves 15 points from cash and gold into stocks, half of it into small-cap value. Tyler wrote in a 2016 follow-up that the value premium “is not central to the portfolio” and that small-cap blend or a broad international fund would also work.⁴

The Cockroach is a different kind of thing. Mutiny describes five sectors (stocks, income, long volatility, commodity trend, and a “fiat hedge” of gold plus bitcoin and ether futures) and allocates a risk budget to each: 23% to each of the first four and 8% to the fiat hedge. Each sector “may be individually levered between 1x and 4x” to hit that budget.⁵ A risk budget is a share of the portfolio’s volatility. The dollar weights and notional exposures that produce it are different numbers, and Mutiny does not publish them. The Cockroach whitepaper gives no weights at all.⁶

The Cockroach column is a share of risk, so it cannot be compared line by line with the dollar weights in the other two columns. A 23% risk budget in long volatility might take only a few percent of capital in options, while 23% in bonds might take more than 100% of capital in Treasury futures.

The theory: three ideas behind all three

Lower volatility compounds closer to the average return

A portfolio’s compound growth rate is roughly its average annual return minus half its variance:

Our backtest shows the approximation working. From 1972 through 2025, 100% US stocks averaged 8.6% a year after inflation with 17.6% volatility. Half of 17.6% squared is 1.5 points, which predicts a compound rate of 7.0%, and that is what it earned. The Golden Butterfly averaged 6.4% with 8.4% volatility, which predicts 6.1%, again what it earned. So the Butterfly lost about 0.3 points a year to volatility and stocks lost about 1.6. That narrowed a 2.2-point gap in average returns to a 0.9-point gap in compound returns. It did not close it.

That is the limit of the smoothness argument. Lower volatility helps compounding, but a portfolio with a much lower average return still compounds slower. The diversification return guide works through the same math and how much rebalancing adds.

Gold behaved more like crisis insurance than an inflation hedge

Browne assigned gold to inflation. The evidence for that is weak over the horizons investors actually hold. Erb and Harvey found that “over practical investment horizons, gold is an unreliable inflation hedge.”⁷ In our data, gold returned 14.8% a year after inflation from 1972 through 1981 and then lost 4.8% a year after inflation over the next 20 years, 1982 through 2001, while inflation stayed positive throughout. The inflation hedges guide covers why TIPS and I bonds do the job gold is usually credited with.

Gold’s case in these portfolios is as a diversifier. It has no cash flow and no agreed valuation model, but from 1972 through 2025 its annual real returns had a correlation of -0.18 with US stocks and -0.13 with long Treasuries in our data. The ablation results below show that removing it deepened the worst drawdown of both portfolios.

Long Treasuries carry 20-year duration risk

A bond’s price falls by roughly its duration times the rise in yields:

A new 20-year Treasury at the September 2026 average yield of 5.4% has a modified duration of about 12, so a one-point rise in yields cuts its price about 11%.⁸ That is the point of holding it: when a recession pulls yields down, long bonds rise sharply. It also means a quarter of the Permanent Portfolio is a large bet against rising rates. Our 20-year bond series lost 23.6% in 2022. The stock-bond correlation that makes long bonds a hedge has changed sign before. Campbell, Sunderam, and Viceira document that the bond-stock covariance “moved considerably over time,” and was negative in the 2000s.⁹ The bonds guide explains when bonds hedge stocks and when they fall with them.

Four regimes are a model of the economy

All three portfolios rest on the idea that the economy cycles through four environments (growth, recession, inflation, deflation) and that each asset has a home in one of them. The model is useful, but the regimes are not equally likely or cleanly separated, and what moves prices is the surprise relative to expectations. The Cockroach whitepaper makes this choice explicit: its historical regimes were “normalized so that each of these combined regimes occurs approximately 25% of the time” over 1969 to 2023.⁶ That is a research design choice. Equal weights across regimes are a way to avoid forecasting, and they rest on no evidence that the regimes are equally probable.

Our backtest, 1972 through 2025

We built every sleeve from primary data, annually, after inflation:

• US stocks: the Fama-French US market return, and small-cap value from the Fama-French small high book-to-market portfolio.¹⁰
• Long Treasuries: a synthetic 20-year bond rebuilt each month from Federal Reserve H.15 yields. The Fed did not publish a 20-year yield from 1987 to 1993, so the 30-year yield fills that gap.⁸
• T-bills: the one-month Treasury bill return from the same French library, used for both the Permanent Portfolio’s cash and the Butterfly’s short-term Treasuries.
• Gold, the 10-year Treasury used in the 60/40, and CPI inflation: Damodaran’s annual series. Its gold returns match LBMA year-end prices.¹¹

We start in 1972 because the US closed the gold window in August 1971. Before that the gold series is a government-set price, and in 1933-34 it shows returns of 27% and 32% that were really a devaluation of the dollar. That is why the risk metrics guide, which runs from 1928 with 10-year bonds, reports different Permanent Portfolio numbers.

As a check, the same method run from 1970 gives average real returns of 4.9% for the Permanent Portfolio and 6.3% for the Golden Butterfly, against the 5.1% and 6.3% Portfolio Charts reports for that window from its own independent reconstruction.¹² The portfolios are rebalanced every December. Nothing here includes fund fees, trading costs, or taxes (taxes get their own section). The academic small-cap value portfolio in particular earned more than any fund you could have bought.

Over the full period the Golden Butterfly beat the 60/40 on return and had well under half its drawdown. The Permanent Portfolio had the smoothest ride of the four, earned the least, and ended with about two-thirds of the 60/40’s real wealth. All-stock investors ended with the most, after a 52% real decline and 13 years below a prior peak. Drawdowns measured on year-end values miss declines that recover within a year, so peak-to-trough losses measured monthly would be at least as deep for every portfolio.

Before and after publication

A backtest that includes the years a portfolio was designed around is partly a description of those years. Browne introduced the Permanent Portfolio in his newsletter in late 1977, near the end of that inflationary decade, and published the equal-quarters version in 1987.² Tyler published the Golden Butterfly in September 2015, so 2016-2025 is its first decade out of sample.⁴ Splitting the record shows how much depended on the 1970s:

The 1970s are where the defensive portfolios won outright: positive real returns while stocks and the 60/40 lost purchasing power for a decade. From 1982 on, the 60/40 beat both, by 2.4 points a year over the Permanent Portfolio. Since the Golden Butterfly was published it has roughly matched the 60/40 and trailed all-stock portfolios by 5.5 points a year, while US large caps ran and small-cap value lagged. That is consistent with the insurance framing: the defensive portfolios paid off in the one decade that punished stocks and bonds together, and cost money in most of the others.

These results match the academic work on the Permanent Portfolio. Anderson, Marshall, and Miao found it did not consistently beat stock or stock-bond portfolios on raw return but delivered better risk-adjusted returns and smaller downside losses.¹³

What each sleeve contributed

To see which sleeves drove the results, we removed one at a time and put its weight in T-bills, then reran 1972-2025:

The ablations show three things.

• Small-cap value accounts for most of the Butterfly’s lead. The Fama-French small-cap value portfolio returned 10.2% a year after inflation against 7.0% for the total market. Swap it for more total market and the Butterfly returns 5.3%, just below the 60/40’s 5.4%, with a worst drawdown of 18%. Since 2016 small-cap value returned 7.6% a year against the market’s 11.2%. If you do not expect a value premium, expect the Butterfly to behave like a Permanent Portfolio with more stocks. The small-cap value guide covers the evidence for and against.
• Gold reduced drawdowns more than any other sleeve. Removing it cut returns by 1.4 to 1.6 points and deepened the worst drawdown in both portfolios, to 29% in the Butterfly. A Butterfly with 10% gold (the other 10% in total market) earned the same 6.1% but had a 24% worst drawdown. Much of gold’s return arrived in 1972-1981 and again since 2000; its contribution depends on whether the next 50 years contain episodes like those.
• Long Treasuries added the least. Removing them cost only 0.5 to 0.7 points and made drawdowns smaller, because the 1970s and 2022 hit them hard. That result leans on two inflation episodes. In a recession with falling inflation, like 2008, when our 20-year series gained 26%, long bonds were the best asset in the portfolio.

One more ablation: a Permanent Portfolio holding 10-year instead of 20-year Treasuries returned 4.4% instead of 4.7%, with a slightly smaller drawdown. Duration added a little return over a period in which yields mostly fell.

The 6% withdrawal rate comes from the 1970s windows

Portfolio Charts reports a 30-year safe withdrawal rate of 6.0% for the Golden Butterfly and 5.5% for the Permanent Portfolio, against 4.1% for a 60/40 and 3.8% for total stock market.^{3, 1} Our reconstruction, whose windows start in 1972 rather than 1970, finds similar numbers. Taking a fixed real withdrawal each year that would have run the portfolio to exactly zero after 30 years, across the 25 starting years from 1972 to 1996:

The defensive portfolios raised the floor. In the median case, the 60/40 and all-stock portfolios supported more spending than either. The Butterfly supported more than the 60/40 only for retirements that began from 1972 through 1978; for every start from 1980 on, the 60/40 supported more. Those 1970s starts are the cases a retiree most needs to survive, and even the Butterfly’s worst window beat the 60/40’s worst. That is a real benefit, and it is the main reason to consider these portfolios for early retirement. See sequence-of-returns risk for why the first decade decides most retirement outcomes.

Three cautions keep us from quoting 6% as a planning number. The 25 windows overlap by up to 29 years, so they amount to a few independent observations, not 25. They come from one country whose stock market did unusually well; Anarkulova, Cederburg, and O’Doherty estimate a 12.1% chance that a 30-year stock investor loses to inflation once you include 39 developed markets.¹⁴ And the sample contains the best period for gold in modern history and a 40-year fall in interest rates. The safe withdrawal rate guide covers how to plan with these numbers.

Rebalancing bands changed the result

Browne’s own rule in Fail-Safe Investing was a band: restore the 25% weights only when a sleeve is worth less than 15% or more than 35% of the portfolio.² Checking the bands once a year, that version returned 5.4% a year after inflation against 4.7% for annual rebalancing, and it beat annual rebalancing in every sub-period we tested. It triggered 14 times in 54 years. Letting winners run longer held more in whatever was trending, which raised the average stock weight to about 28%. The gap is a reminder that small rule choices move backtests, and that published results for “the” Permanent Portfolio depend on which rule the author used. The rebalancing guide covers band rules in general.

2022: when stocks and bonds fell together

In 2022 US stocks lost 19.9% and our 20-year Treasury series lost 23.6%, while gold returned 0.5% in a year of 6.5% inflation. The Permanent Portfolio lost 10.4% (15.8% after inflation) and the Golden Butterfly 9.6% (15.0%), against 19.1% (24.0%) for the 60/40. Gold and cash held the defensive portfolios up while both main assets fell. Neither came close to breaking even. In a year that hit long bonds and stocks together, a portfolio with a quarter in long bonds still lost money.

Taxes cost these portfolios more than they cost stocks

Every sleeve outside stocks is taxed less favorably in a brokerage account:

• T-bill and Treasury interest is taxed as ordinary income every year, though it is exempt from state and local income tax.¹⁵
• Long-term gains on collectibles, which include physically backed gold trusts such as GLD, are taxed at up to 28% instead of the usual maximum of 20%.^{16, 17} For someone in a bracket below 28%, the collectibles rate is their ordinary rate, which is still above the 15% long-term rate.
• Annual rebalancing sells the sleeve that rose, which realizes gains each year instead of deferring them.

To size this, we ran each portfolio in a taxable account from 1972 through 2025 at today’s federal rates for two investors. One is in the 24% bracket (15% on dividends and long-term gains, 24% on gold gains). The other is in the 35% bracket and pays the 3.8% net investment income tax.¹⁸ Taxes were paid from the portfolio each year and remaining gains were taxed at the end. State taxes are left out; including them would add drag to stocks and gold but not to Treasury interest, which is state-exempt.

Taxes took about 1.4 points a year from the Permanent Portfolio and Golden Butterfly in the 24% bracket and about 2 points in the top case, roughly twice the drag on all-stock portfolios. Part of the drag is tax on inflation: the 1970s interest and gold gains were taxed in nominal dollars. Asset location shrinks the drag: holding the Treasuries and gold in IRAs or 401(k)s and the stocks in the taxable account keeps the same allocation and pays less tax. The model is stylized (it treats every sale as long-term and uses one set of rates for 54 years), so read the drag as an order of magnitude.

The Cockroach as a fund you can buy

The Cockroach builds on the Permanent Portfolio and adds two sleeves the passive portfolios lack: commodity trend following and long volatility. There is real research behind trend following, which the managed futures guide covers, and tail hedging can pay off in crashes, which the tail-risk hedging guide covers. Mutiny’s whitepaper also diversifies within each sleeve across several managers and signals, which it calls “fractal diversification.”⁶

The evidence for those building blocks does not tell you whether this fund delivers them after costs. As of October 2026, here is what Mutiny publishes:^{5, 19}

• A 1% management fee and a 10% incentive fee, a $100,000 minimum, and eligibility limited to accredited investors.
• Monthly redemptions with 10 business days’ notice, with up to 10% of an investment held back until the next calendar year to finish the fund’s audit, which Mutiny says is not typical.
• A K-1 tax form typically around August, which means filing an extension, and no planned distributions to cover the tax on your share of the fund’s gains.
• No public performance record. The whitepaper’s charts are labeled illustrative, and returns are available to qualified investors on request.

The whitepaper is a statement of the manager’s philosophy, and some of its arguments need care. It says standalone tail-risk hedging is “a long-term losing strategy” and that long volatility is “the most difficult to evaluate historically.” It suggests 1.5x to 3x leverage, and it equates a bond’s duration with its maturity, which holds only for zero-coupon bonds.⁶ Mutiny’s site also says that “it is not possible, to our knowledge, to include active managers in an ETF structure.” The SEC’s 2019 ETF rule provides exemptions “for both index-based ETFs and actively managed ETFs” that disclose their holdings daily.²⁰ The narrower point holds: an ETF cannot easily hand pieces of its portfolio to outside private managers, which is how Mutiny runs its trend and volatility sleeves.

A DIY investor can add trend following through managed-futures ETFs, as the Hungry Caterpillar portfolio does. Rebuilding a diversified long-volatility sleeve is harder: it takes option expertise, and a simple put-buying program is the steady-loss strategy the whitepaper itself warns about.

PRPFX holds a different portfolio

The Permanent Portfolio fund (PRPFX) uses Browne’s name but not his allocation. Its targets are 20% gold, 5% silver, 10% Swiss franc assets, 15% real estate and natural resource stocks, 15% aggressive growth stocks, and 35% dollar assets, and it charges 0.79% a year for its institutional share class.²¹ At the end of June 2026, 22% of the fund was in corporate bonds and 11% in Treasuries.²² It returned -5.49% in 2022, a better year than our 25/25/25/25 portfolio had. Results quoted for PRPFX tell you about that fund; they are not a test of Browne’s design.

Who each portfolio is for

Behavior is the argument for all of these. A portfolio you hold through a 50% decline beats a better one you sell at the bottom, and the reverse holds too: an investor who abandons a Butterfly after trailing the S&P 500 for a decade ends up worse than one who held a plain index fund.

Our recommendation

• Default: a low-cost global stock and bond portfolio sized to your risk tolerance. It is cheaper, more tax-efficient, and does not depend on gold or small-cap value repeating their history.
• If you are retired or close to it and want a floor: the Golden Butterfly is a reasonable structure. Hold the Treasuries and gold in tax-advantaged accounts. Keep the small-cap value sleeve only if you expect the value premium to persist; without it, expect returns closer to a 60/40 with a smaller drawdown.
• If preservation matters more than growth: the Permanent Portfolio does what it says, at a cost of about 2.4 points a year against a 60/40 since 1982 in our data.
• The Cockroach: treat the published framework as a set of ideas you can borrow (diversify the defensive side, add trend) and the fund as a private investment that needs the same due diligence as any hedge fund.

The exact weights in any of these are less important than deciding which bad outcome you most need to avoid. The DeMiguel, Garlappi, and Uppal finding that simple equal weights are hard to beat out of sample is a reason to keep the allocation simple. It does not show that 25% or 20% is the right weight for any sleeve.²³

Key takeaways

• In 1972-2025 US data, the Golden Butterfly earned 6.1% a year after inflation with a 15% worst drawdown. A 60/40 earned 5.4% with a 38% drawdown, and the Permanent Portfolio 4.7% with 18%.
• Small-cap value supplied most of the Butterfly’s edge. Without it the Butterfly earned 5.3%, using an academic index with no fees.
• The defensive portfolios won the 1970s and trailed after. From 1982 the 60/40 beat the Permanent Portfolio by 2.4 points a year.
• They raised the worst-case withdrawal rate. The Butterfly’s worst 30-year window supported 6.4% a year in our data against 4.3% for a 60/40, but its median was lower.
• Taxes cost the defensive portfolios about twice what they cost stocks. In a taxable account that was about 1.4 to 2 points a year, so the Treasury and gold sleeves belong in tax-advantaged accounts.
• The Cockroach publishes risk budgets but no returns. Judge the fund on audited net results, which are available only on request.

Frequently asked questions

Is the Golden Butterfly better than the Permanent Portfolio?

In our 1972-2025 US backtest it earned 1.4 points a year more after inflation with a slightly smaller worst drawdown, because it holds 40% in stocks instead of 25%. Most of that gap came from its small-cap value sleeve. Without it, the Butterfly beat the Permanent Portfolio by 0.6 points.

Why does the Permanent Portfolio hold 25% in gold?

Browne assigned gold to inflationary periods. Over typical investment horizons gold has been an unreliable inflation hedge, but in our data it was the sleeve that most reduced drawdowns. Removing it lowered the Permanent Portfolio’s return by 1.6 points a year and deepened its worst drawdown from 18% to 23%.

What is the Cockroach portfolio allocation?

Mutiny publishes a risk budget, not capital weights: 23% each to stocks, income (bonds and carry), long volatility, and commodity trend, and 8% to a fiat hedge of gold and crypto futures, with each sector levered between 1x and 4x. You cannot reproduce it by putting 23% of your money in each of four ETFs.

Can I build the Cockroach with ETFs?

You can approximate parts of it. Stock, Treasury, gold, and managed-futures ETFs cover four of its sleeves without leverage. A diversified long-volatility sleeve has no simple ETF equivalent, and Mutiny’s version uses outside managers and leverage. Any ETF portfolio called a Cockroach is a different portfolio from the fund.

Do these portfolios work outside the US?

Our backtest and Portfolio Charts’ headline figures use US assets. Gold and Treasuries are dollar-denominated, and investors in other currencies get different results. The Butterfly also holds only US stocks; a global stock sleeve is a reasonable substitute that Tyler himself mentions.

Related guides

• The Hungry Caterpillar Portfolio is a global descendant of all three, with a trend sleeve and a portfolio builder.
• Do You Need Managed Futures? covers the trend sleeve the Cockroach adds.
• The Risk Parity Reality Check explains risk budgets and leverage.
• Diversification Return works through the rebalancing math.
• The Best Inflation Hedges compares gold with TIPS and I bonds.
• Are Bonds Still Good Diversifiers? covers when long Treasuries hedge stocks.
• Tail-Risk Hedging covers the cost of the Cockroach’s long-volatility sleeve.

Sources

Backtest figures are the author’s calculation from the data sources below, reproducible with scripts/analyze_all_weather_portfolios.py in the Summitward repository. Fund terms were read from issuer documents in October 2026.

1. Portfolio Charts. Permanent Portfolio. Allocation and US 1970-2025 statistics (mean real return 5.1%, 30-year safe withdrawal rate 5.5%).
2. Browne, H. (1987). Why the Best-Laid Investment Plans Usually Go Wrong. New York: Morrow. Equal-quarters Permanent Portfolio; dates the concept to his late-1977 newsletter. Browne, H. (1999). Fail-Safe Investing. New York: St. Martin’s Press. The 15% and 35% rebalancing bands.
3. Portfolio Charts. Golden Butterfly. Allocation and US 1970-2025 statistics (mean real return 6.3%, 30-year safe withdrawal rate 6.0%, deepest drawdown 18%).
4. Tyler, Portfolio Charts. Catching a Golden Butterfly (September 22, 2015) and The Theory Behind the Golden Butterfly (April 18, 2016).
5. Mutiny Fund. Cockroach Strategy. Risk budget by sector, leverage range, fees, minimum, redemption terms. Read October 2026.
6. Pearson, T., with Buck, J. (2024). The Cockroach Approach. Mutiny Fund whitepaper. Regime normalization (p. 30), illustrative charts (pp. 61-68), construction (pp. 66-79), duration (p. 71).
7. Erb, C. B., & Harvey, C. R. (2013). The Golden Dilemma. NBER Working Paper 18706; published in the Financial Analysts Journal 69(4).
8. Board of Governors of the Federal Reserve System. H.15 Selected Interest Rates. Monthly 10-, 20-, and 30-year constant-maturity Treasury yields, retrieved through DBnomics.
9. Campbell, J. Y., Sunderam, A., & Viceira, L. M. (2017). Inflation Bets or Deflation Hedges? The Changing Risks of Nominal Bonds. Critical Finance Review 6(2), 263-301.
10. Kenneth R. French. Data Library. Fama/French 3 Factors and 6 Portfolios Formed on Size and Book-to-Market (2 x 3), annual returns.
11. Aswath Damodaran. Historical Returns on Stocks, Bonds and Bills. Annual gold, 10-year Treasury, and CPI series, 1928-2025.
12. Portfolio Charts. Data Sources.
13. Anderson, H., Marshall, B. R., & Miao, J. (2014). The Permanent Portfolio. Applied Financial Economics 24(16), 1083-1089.
14. Anarkulova, A., Cederburg, S., & O’Doherty, M. S. (2022). Stocks for the Long Run? Evidence from a Broad Sample of Developed Markets. Journal of Financial Economics 143(1). Summary by the authors.
15. TreasuryDirect. Treasury Bonds. State and local tax exemption.
16. Internal Revenue Service. Topic No. 409, Capital Gains and Losses.
17. SPDR Gold Shares. GLD FAQs (January 2026). Tax treatment as a collectible.
18. Internal Revenue Service. Net Investment Income Tax. 3.8% on interest, dividends, and capital gains above $200,000 (single) or $250,000 (married filing jointly) of modified adjusted gross income.
19. Mutiny Fund. Due Diligence Questionnaire (updated January 1, 2025). K-1 timing, distributions, holdback, gold and crypto futures holdings.
20. U.S. Securities and Exchange Commission (2019). Exchange-Traded Funds, Release Nos. 33-10695; IC-33646 (Rule 6c-11).
21. Permanent Portfolio Family of Funds. Permanent Portfolio Summary Prospectus. Target allocations, expenses, annual returns.
22. Permanent Portfolio Family of Funds. PRPFX Fact Sheet (June 30, 2026).
23. DeMiguel, V., Garlappi, L., & Uppal, R. (2009). Optimal Versus Naive Diversification: How Inefficient Is the 1/N Portfolio Strategy?. Review of Financial Studies 22(5), 1915-1953.