Backtest · May 2007 – Oct 2026 · monthly · TradingView data

All-Weather India

Ray Dalio's four-seasons portfolio rebuilt with Indian instruments. Set your own weights and see how it behaves against the Nifty 50 and a 60/40 mix.

Start here

Ray Dalio's idea: the economy has four seasons, and no one knows which is next. Growth rises, growth falls, prices rise, prices fall. So hold something that does well in each, and sleep through all of them.

The India version: 30% stocks (25% Nifty 50 and 5% realty), 55% government bonds (40% long, 15% medium), 7.5% gold and 7.5% commodities, reset to these weights once a year.

Growth close to a standard 60/40 mix, with far less risk.

Against a standard 60% Nifty / 40% bonds mix, All-Weather India grew 0.8 points a year less (9.6% against 10.3%), but its ups and downs were 39% smaller and its worst fall was 47% shallower.

	All-Weather India	60/40 mix	100% Nifty
Growth per year	9.6%	10.3%	11.4%
Ups and downs (volatility)	7.4%	12.1%	20.0%
Worst fall	16.7%	31.8%	54.6%
Growth for each point of risk	1.29	0.85	0.57
Return for the risk (Sharpe)	0.48	0.36	0.27

May 2007 to Oct 2026. The price: ₹1 lakh became ₹5.93L, against ₹6.79L in 60/40 and ₹8.21L in the Nifty. It gives up some growth to avoid the big crashes. Growth for each point of risk is yearly growth divided by volatility.

Below: move the sliders to try your own mix. Each chart has a short note beside it. The scoreboard at the very end ranks everything.

Allocation

Rebalance Risk-free for Sharpe 6.0%

Growth of ₹1 lakh

Drawdown from previous peak

Scorecard

Calendar years

Stress windows

How the sleeves move together

What is under the hood

Sample: 234 monthly returns, May 2007 to Oct 2026. November 2026 is left out because it is only a few days old.
Equity: NIFTYBEES price, plus an assumed 1.3% a year in dividends because the price series leaves them out.
Real estate: Nifty Realty index, price only. It tracks property developers' shares, not rental REITs, and moves with the Nifty (correlation 0.82). Embassy REIT only has data from 2019.
Bonds: built from the India 10-year and 5-year yields. Each month earns the yield divided by 12, minus duration times the change in yield, minus a fitted residual that covers fees and any mismatch between the ETF and the bond yield. Duration and residual were fitted to the real LTGILTBEES and GILT5YBEES ETFs over May 2022 to Oct 2026: duration 6.2 and 0.5 points a year for the long sleeve, duration 3.75 and about zero for the mid sleeve. Monthly correlation with the ETFs is 0.94 and 0.86. This is an in-sample fit on 53 months. Fitted on each half separately, the long duration comes out at 5.5 and 6.7, so a duration of 7 is not ruled out. Using the fit back to 2007 is an assumption, and the 2008 result leans on it. An earlier model (duration 7 and 4, convexity, no residual) would show the All-Weather mix at 9.9% a year instead of 9.6%.
Gold: GOLDBEES in rupees, so it includes the rupee falling from 41 to 95 per dollar.
Commodities: the DBC fund priced in dollars and converted to rupees. Indian investors cannot buy it directly. It returned 1.4% a year in dollars and 5.9% in rupees, so most of that comes from the currency. The price series leaves out the fund's interest income.
Not included: taxes and trading costs. NIFTYBEES, GOLDBEES and DBC prices are already net of their own fees. The Realty index has no fee deducted. Fees on the bond sleeves only show up through the fitted residual.
Scorecard: Sharpe is (CAGR minus 6%) divided by volatility. Drawdown uses month-end values, so intramonth lows are missed. The inverse-volatility weights use the whole period's volatility, so they could not have been known in 2007.

Who wins what

Four mixes, using only what was knowable in 2007 (the weights, not the instruments), over the same 234 months. All-Weather India is outlined in blue wherever it comes first.