All Weather, Rebuilt for Canada

Ray Dalio’s portfolio for every economic season, taken back to its first principles for a Canadian investor in 2026 — twenty years of backtests in Canadian dollars, and the decisions that shape any version you build

A first-principles guide to the All-Weather portfolio for Canadians in 2026: the four economic seasons, balancing risk rather than dollars, why stocks and bonds fell together in 2022, what the popular 30/40/15/7.5/7.5 mix did in Canadian dollars since 2007, an equal-risk version built without hindsight, and the nine decisions a Canadian has to make: return against loss, fixed or risk-balanced weights, duration or leverage, currency, inflation protection, funds, accounts, rebalancing, and building or buying. Arithmetic, not advice.
Author

Behzad Samadi

Published

October 28, 2026

A paper-craft cabin on a hill under a navy and amber umbrella, with the four seasons around it: summer fields, autumn maples, winter pines in snow, and spring rain over a lake. Title: All Weather, Rebuilt for Canada. First principles, tested on twenty years.

Most portfolios are built for one kind of weather. A portfolio of stocks does well when the economy grows and badly when it shrinks. A portfolio of bonds does well when inflation falls and badly when it rises. Ray Dalio, the founder of Bridgewater Associates, asked a different question in the 1990s: what would you hold if you had no idea which weather was coming? His answer, the All Weather portfolio, spreads its risk evenly across every economic season instead of betting on one.

Most people know it as a simple recipe: 30% stocks, 40% long-term US government bonds, 15% medium-term ones, 7.5% gold and 7.5% commodities. That recipe is a 2014 simplification, written for Americans, from a book by Tony Robbins. Dalio’s own comment on it was “It’s not really that simple.” It was built in a world of falling interest rates and of stocks and bonds that moved in opposite directions. In 2022 that world ended: stocks and bonds fell together, and the recipe lost about a fifth of its value.

This article goes back to the principles underneath the recipe and tests the recipe on twenty years of data in Canadian dollars. It does not end with a portfolio to copy. There is no single right All Weather; the principles leave a handful of real choices open, and each one changes how the portfolio behaves. The aim here is to lay those choices out, with what each trades off and what the evidence says, so that whatever you build, you know why.

NoteIn short
  1. The idea holds up; the recipe is dated. Balance risk across four economic seasons and don’t try to predict which is next. The popular 30/40/15/7.5/7.5 mix doesn’t actually do that: about half its risk sits in long-term US government bonds.
  2. In Canadian dollars, since 2007, the recipe earned about 6–7.5% a year with far smaller losses than stocks in 2008, but it lost 11–19% in 2022, when stocks and bonds fell together. Global stocks earned about 9% a year with a 42% fall along the way.
  3. For a Canadian, the US dollar is part of the hedge. In the 2008 crash the US dollar rose about 30% against the Canadian dollar, so US bonds held unhedged gained 63% in Canadian dollars that year.
  4. Weights set by the principle instead of the book (equal risk from each part, using only the past three years’ data each January) matched the recipe’s 6.6% a year since 2010 with less than half the worst loss. One period; it shows what the principle does, not what it will do.
  5. 2026 is different: bonds pay 4–5% again, Canadian inflation is above 3% (the level above which stocks and bonds usually move together), Ottawa no longer issues inflation-protected bonds, and Bridgewater now sells its own version as an ETF.
  6. There is no single answer. Nine decisions shape any version you build, from how much return you will give up for a smaller fall to whether to buy Bridgewater’s own fund. This article lays them out rather than choosing for you.
WarningThis is arithmetic, not advice

Nothing here knows your goals, your timeline or how you would feel watching a portfolio fall. The backtests below are exact for their inputs, and the inputs are published with the code, but the past twenty years are one sample of history among many possible ones. Every rate and price is dated October 2026.

The first principles

1. Prices move on surprises, not on news

Bridgewater’s starting point is that every investment earns, over time, roughly the return on cash plus a reward for taking risk. What moves prices from day to day is not good or bad news as such, but news that is better or worse than what was already expected. A strong economy that everyone expected does little to stock prices; one that surprises does.

Nobody can predict surprises reliably. So instead of trying, All Weather asks: which kinds of surprise matter most, and can a portfolio be built so that it doesn’t depend on which one comes?

2. Two surprises matter most: growth and inflation

Dalio reduces the economy to two dials, each of which can turn out higher or lower than expected. That gives four seasons, and each season favours different assets:

Season What happens What tends to do well
Growth higher than expected Output and profits beat forecasts Stocks, commodities, corporate bonds
Growth lower than expected The economy slows; recession risk rises Government bonds, inflation-protected bonds, cash
Inflation higher than expected Prices climb faster than forecast Commodities, gold, inflation-protected bonds
Inflation lower than expected Prices rise slower than forecast, or fall Government bonds, stocks

Bridgewater puts a quarter of the portfolio’s risk in each season. Some assets appear in two rows, which is why the weights have to be worked out from risk, not assigned one asset per season.

3. Balance risk, not dollars

This is the principle most people miss. A portfolio that is half stocks and half bonds by dollars is not balanced, because stocks swing far more than bonds. In our data, a Canadian with half in global stocks and half in a broad Canadian bond fund had 79% of the portfolio’s ups and downs coming from stocks. Bridgewater puts the figure for a typical 60/40 pension fund at about 90%. Such a portfolio is really a stock portfolio with a cushion.

How much each asset swings (annual volatility, 2007–2026, in US dollars):

Asset Volatility
US stocks 15.6%
US government bonds, 20+ years 13.8%
US government bonds, 7–10 years 6.6%
Gold 17.5%
Broad commodities 18.8%

To give a calm asset the same weight in risk as a wild one, you need more of it in dollars.

4. Make the calm assets count: borrow, or go long

There are two ways to get enough risk out of bonds to balance stocks. Bridgewater’s real fund borrows: it holds bonds with leverage, for example $2 of a ten-year bond for every $1 of capital. The retail recipe can’t borrow, so it does the other thing: it holds very long bonds, whose prices swing almost as much as stocks.

That choice has a cost. Measured on the same data, the retail recipe gets about 48% of its risk from long-term US government bonds and only 33% from stocks; gold and commodities together carry about 11%. It is not balanced across four seasons. It is a large bet that long-term interest rates won’t rise sharply, and in 2022 they did.

5. Whether stocks and bonds move together depends on inflation

The recipe relies on long bonds rising when stocks fall. From 2000 to 2020 they usually did: monthly US stock and long-bond returns had a correlation of about −0.30. But that is not a law of nature. From 1986 to 1999 it was +0.27, and since 2021 it has been +0.54.

The reason is the second dial. When the main surprise is growth, stocks and bonds move in opposite directions. When the main surprise is inflation, both suffer, because rising prices push up interest rates, which lowers bond prices and the value of future profits. Bridgewater said so in its own materials in 2009. Man Group’s research finds that when inflation is above 3%, the stock-bond correlation has been negative only 2% of the time. In September 2026 Canada’s inflation rate was 3.1%.

6. Don’t predict; rebalance

All Weather is meant to be held, not timed. Once a year (or when a part drifts far from its target), you sell some of what has risen and buy some of what has fallen. That keeps the risk where you meant it to be. It is mainly a discipline, not a reliable source of extra return. Vanguard finds little difference between rebalancing monthly, quarterly or yearly, except in cost.

Diversification has limits too. Dalio says 15 good, uncorrelated sources of return cut risk by about 80%. The arithmetic gives about 74% even if they are perfectly uncorrelated, and real assets are not. A portfolio of five ETFs gives you perhaps three or four truly independent bets.

7. For a Canadian, currency is an asset too

A Canadian who holds US-dollar assets without currency hedging also holds US dollars, and the US dollar tends to rise against the Canadian dollar in a crisis. The Canadian dollar is tied to oil and to the world economy; when both fall, it falls. Bank of Canada data:

Episode US dollar against the Canadian dollar
July 2008 to the March 2009 market low +29.7%
The COVID crash, February to March 2020 +9.6%
2022 +6.8%
2025 −4.7%

So in 2008, long-term US government bonds returned about 34% in US dollars, and about 63% in Canadian dollars. The same exposure costs you when the Canadian dollar strengthens, as in 2025. Hedged and unhedged versions of the same bond fund behave like different assets.

What equal risk would have done

If the principle is “equal risk from each part”, we can let the data set the weights instead of a book. Each January we looked only at the previous three years of monthly returns of five building blocks (global stocks, Canadian long bonds, Canadian government bonds, gold and commodities, all in Canadian dollars) and solved for the weights at which each one contributes the same share of the portfolio’s risk. No hindsight: each year’s weights use only data that existed then. Three years of data are needed first, so this runs from 2010.

2010 to September 2026 Return a year Volatility Worst fall Worst year 2022
Equal-risk version 6.6% 5.7% −6.1% −2.4% (2022) −2.4%
The recipe, Canadian version 6.6% 6.8% −14.2% −11.0% (2022) −11.0%
60% global stocks, 40% bonds 8.2% 7.6% −15.3% −10.5% (2022) −10.5%

The weights it chose moved with the markets. In January 2026, from the three years before:

Part Weight chosen for 2026
Global stocks 17%
Canadian long bonds 17%
Canadian government bonds 30%
Gold 14%
Commodities 22%

Three things stand out. Because it balances risk, it holds much more gold and commodities than the recipe (36% against 15%) and much less in long bonds, which is exactly what protected it in 2022. It earned the same as the recipe with less than half the worst loss. And it still trailed a plain 60/40 portfolio in a period that was mostly kind to stocks. This is one period, chosen by data availability, and it happened to include a commodity boom. It shows what the principle does; it does not promise what it will do.

What is different in 2026

Bonds pay again. On 27 October 2026 a long Government of Canada bond yielded 4.12%, and a fund of them (ZFL) 4.27% to maturity. The 30-year US Treasury yielded 5.30%. A bond’s starting yield is the best simple guide to what it will earn if held to maturity, so the bond side of the portfolio starts from a better place than it did when yields were near their lows.

Real yields are positive. A ten-year US inflation-protected bond paid 2.64% above inflation, and Canada’s long Real Return Bond 1.87%. In 2020–21 the US ten-year real yield was below zero.

Inflation is back above 3%, and the Bank of Canada raised its rate on 28 October to 2.50%, its first increase in about three years, worried that high energy prices would spread. That is the season in which stocks and bonds tend to fall together, and in which commodities and short-term, inflation-linked assets matter most.

Ottawa no longer issues inflation-protected bonds. The federal government stopped issuing Real Return Bonds on 3 November 2022, citing low demand. The existing ones still trade, and an ETF (XRB) holds them, but the supply is shrinking. Canadians who want this part now mostly use short-term US TIPS ETFs, which protect against US inflation, not Canadian.

Gold has run, and stumbled. Gold rose about 25% in 2024 and 67% in 2025, reached a record near US$5,590 an ounce in January 2026, and has since fallen to about US$4,030. Broad commodities have risen about a third in 2026.

Bridgewater now sells its own version. Since March 2025, the State Street Bridgewater All Weather ETF (ALLW, US-listed, 0.85% a year) has held Bridgewater’s actual design. On 30 September 2026 it held, for every $100, about $71 of government bonds from several countries, $44 of stocks, $37 of US inflation-protected bonds, $23 of commodities and $10 of gold: about $180 of exposure, using leverage, and nothing Canadian. It returned 12.1% from launch to the end of September 2026, against 21.2% for world stocks.

Dalio himself is tilting. In August 2026 Dalio said he would hold fewer bonds, 10–15% in gold and “a bit” of bitcoin, warning of a possible US debt crisis within a few years. That is a prediction, which is exactly what All Weather was designed not to need. If you follow All Weather, you are following the design, not the designer’s latest view.

The decisions that matter

The principles fix the shape of an All Weather portfolio: balance risk across growth and inflation, and don’t predict. They leave the rest to you. These are the choices that change how it behaves, roughly in order of importance.

1. How much return will you give up for a smaller fall?

This is the first decision, and no backtest makes it for you. From 2007 to September 2026, in Canadian dollars:

Return a year Worst fall
Global stocks 8.9% −41.7%
60% stocks, 40% Canadian bonds 6.8% −23.2%
The recipe, Canadian version 5.9% −14.2%

The Canadian version beat global stocks in only 5 of the 19 full years (2007, 2008, 2010, 2011 and 2018). Over long stretches it will look like the wrong choice. The factors that matter:

  • How long until you need the money. A deep fall just before you start withdrawing does lasting damage; the same fall 25 years out may not matter.
  • What you did in the last big fall. A portfolio you abandon at the bottom earns less than a worse one you keep.
  • Whether you can watch it trail a stock index for years without switching.

2. Fixed weights, or weights that balance risk?

  • Fixed weights (the recipe, or any set you choose) are simple and need no calculation. But they balance risk only for the period they were designed on. The recipe puts about half its risk in long bonds.
  • Risk-balanced weights follow the principle directly, as in the equal-risk backtest above. They need a formula and a choice of data window, and they move: commodities ranged from 8% to 28% between 2010 and
    1. A shorter window reacts faster and chases recent calm; a longer one is steadier and slower.
  • In between: keep fixed weights, but check every year or two whether each part’s swings are still roughly equal, and adjust deliberately.

3. How do you get enough out of bonds: long bonds, leverage, or neither?

Bonds only balance stocks if they carry similar risk. There are three ways to handle that, and each fails differently:

  • Long-duration bonds (the recipe’s way). No borrowing, but a large bet on interest rates. In 2022, Canadian long bonds lost 22% and US long bonds 27% in Canadian dollars.
  • Leverage (Bridgewater’s way). Medium-term bonds held at about twice their value, through futures. In Canada the only “return-stacked” fund that does this is RGBM (1.22% a year); US-listed RSSB and NTSX also exist, and Bridgewater’s own ALLW holds about $1.80 of exposure for every dollar. Leverage costs interest and fees, and in a year like 2022, when both stocks and bonds fall, it makes the loss larger.
  • Neither: accept that bonds will carry less risk, and balance with gold and commodities instead. That moves the portfolio’s risk toward the inflation seasons.

In 2026 the starting point matters: long Canadian bonds yield about 4.1%, and a bond’s starting yield is the best simple guide to what it earns if held to maturity.

4. Canadian bonds or US bonds, and should you hedge the currency?

  • Canadian federal bonds match the currency you spend and Canadian inflation. No currency risk.
  • US bonds, unhedged, add the US dollar, which has risen in every recent crisis: US long bonds gained 63% in Canadian dollars in 2008. The cost is bigger swings in normal times and losses when the Canadian dollar strengthens, as in 2025. They also pay more today (US 10-year 4.99% against Canada’s 3.83%).
  • US bonds, hedged, behave like the textbook bond leg, without the currency cushion: in 2022 the hedged BMO long US Treasury fund lost about 31%, the unhedged one about 27%.

One more factor: an all-in-one stock ETF is already 70–80% foreign, so you may hold more US dollars than you think.

5. What protects you in the inflation season?

This is the hardest part for a Canadian in 2026, because Ottawa no longer issues Real Return Bonds. The options, and what research on past US inflation episodes found:

  • Broad commodities. The most reliable hedge: positive in real terms in every inflation episode Man Group studied (about +14% a year on average), and +27% in Canadian dollars in 2022. A drag the rest of the time, and very volatile.
  • Gold. Strong on average in inflations (about +13% a year real) but unreliable: it worked in about two of three episodes. It has also been unusually volatile lately: a record in January 2026, then a fall of about 28% by October.
  • Inflation-linked bonds. Steady but modest (about +2% a year real in inflations). The Canadian ones (XRB) are a shrinking pool; short US TIPS (XSTH) protect against US inflation, not Canadian.
  • Trend following. Funds that follow price trends gained about 27% in
    1. Not part of All Weather, they lose in trendless years, and in Canada they are hard to find as ETFs.

How much to put here is the real decision: the recipe gives it 15%; the equal-risk weights gave it 36% for 2026, after a commodity boom.

6. Which fund for each part?

Within each part, funds differ in what they hold, how big they are, how they are built and what they cost. Options listed in Canada, with fees as published in September 2026:

Part Options Annual cost (MER) What to weigh
Stocks XEQT, VEQT, ZEQT 0.18–0.22% All global and unhedged; 20–30% Canada, far above Canada’s share of world markets
Long government bonds ZFL (federal); ZTL, XTLT (US, unhedged); ZTL.F, XTLH (US, hedged) 0.18–0.23% XLB is cheaper and older but holds provincial and corporate bonds, which add credit risk
Medium-term government bonds ZFM (federal); XGB, VGV 0.09–0.22% XGB and VGV are cheaper but about half provincial
Gold ZGLD, KILO.B, CGL.C, MNT 0.20–0.55% All physical gold; same metal, different fees and custodians
Commodities ZCOM, CCOM ZCOM 0.26% fee (no MER yet); CCOM 0.66% ZCOM always holds a broad index but is a year old; CCOM follows trends and sometimes sits in cash, which changes what it does in an inflation shock
Inflation-linked XRB, XSTH 0.10–0.39% See decision 5

Smaller funds (ZTL, XFLB, ZCOM) can have wider gaps between buying and selling prices; use limit orders.

7. Which account holds which part?

  • Bonds and commodity funds in an RRSP or TFSA first. Their income is taxed as interest or arrives as unpredictable distributions.
  • Canadian-listed funds holding US government bonds avoid US withholding tax in any account, according to BlackRock Canada’s guide.
  • US-listed funds (ALLW, US commodity funds) only in an RRSP, where the treaty removes the 15% US withholding tax; in a TFSA it can’t be recovered. US commodity funds structured as partnerships (DBC, GSG) carry a 10% US withholding on sale proceeds for non-US sellers since 2023.
  • Stocks and gold are the most tolerable parts in a taxable account, where gains are half-taxed.

8. When do you rebalance?

Once a year, or when a part drifts a set amount (five percentage points is common) from its target. Research finds little difference in risk-adjusted results between frequencies, only in cost. In a taxable account, rebalancing can realise gains, and the superficial loss rule denies a loss if you buy the same thing back within 30 days. New contributions are the cheapest way to rebalance: put them into whatever is below target.

9. Build it, or buy it?

  • Bridgewater’s own ETF (ALLW) is the real design: leverage, inflation-linked bonds, several countries, run by the people who invented it. It costs 0.85% a year, trades in US dollars, holds nothing Canadian, belongs in an RRSP for tax reasons, and trailed world stocks by about 9 percentage points in its first 19 months.
  • Building it yourself from Canadian ETFs costs about 0.2–0.25% a year, keeps it in Canadian dollars and can include Canadian bonds, but without leverage or new Canadian inflation-linked bonds it can only approximate the design, and you make (and keep) every decision above.

Who it suits, and who it doesn’t

It may suit you if:

  • a deep fall would make you sell, so a smoother ride is worth less return;
  • you are close to drawing on the money, when a large early loss does the most damage;
  • you want a portfolio you can hold through any news without deciding anything.

It is probably not for you if:

  • you have decades to go and can ignore a 40% fall: over most long periods, stocks have earned more;
  • you would abandon it after a few years of trailing a stock index, which it will do in strong markets;
  • you would hold its bonds outside a registered account.

Questions to answer before you build anything

  1. What is the worst year you could live through without selling?
  2. How much of the risk in what you hold now comes from stocks? (In a 60/40 portfolio, about 90%.)
  3. Fixed weights, or weights that keep each part’s risk equal, and how often will you review them?
  4. Long bonds, leverage, or more gold and commodities to balance stocks?
  5. How much US-dollar exposure do you already have, and do you want more?
  6. How much goes to the inflation season, and in what?
  7. Which account will hold each part?
  8. What is your rebalancing rule, written down?
  9. Would you rather pay 0.85% for Bridgewater’s design, or about 0.2% to make the decisions yourself?

Sources

The backtest code, its results, and research notes recording what was read in full, what came from search summaries, and what could not be verified are in the accompanying folder. Prices: Yahoo Finance, distribution-adjusted, downloaded 28 October 2026. Canadian CPI: Bank of Canada Valet, series V41690973. Canadian real GDP: FRED, NGDPRSAXDCCAQ.