We tested a simple trend-following rule across six US sector ETFs over three years (2023-08-29 to 2026-08-27). The honest result: it lagged SPY on raw return, but its worst drawdown was about half as deep.
The rule
- Universe: XLK, XLF, XLI, XLE, XLV, XLY (six liquid sector ETFs)
- Buy when the 10-day EMA rises above the 30-day EMA and the 14-day RSI is above 50
- Sell when the 10-day EMA falls back below the 30-day EMA
- $100,000 starting capital, 79 closed trades, 40.5% win rate
Results
| Metric | Sector strategy | Equal-weight sector basket | SPY |
|---|---|---|---|
| Total return | 58.57% | 69.44% | 77.04% |
| Annualised return (CAGR) | 16.66% | 19.27% | 21.01% |
| Max drawdown | -9.81% | -17.51% | -18.20% |
The account grew from $100,000 to $158,574. It gave up some upside to SPY and to simply holding the six ETFs, in exchange for a much shallower worst decline. Whether that trade is worth it depends on how much drawdown you can live with.
Full write-up
Methodology, walk-forward check, monthly returns, drawdown analysis and the code are in the original article on quantstr.at:
Can a Simple Trading Rule Reduce Stock-Market Losses? We Tested It on 6 Sector ETFs
Backtests are historical simulations, not investment advice.
Top comments (3)
Return per unit of drawdown does favour the rule here (16.7/9.8 is about 1.7 against roughly 1.15 for SPY), but part of a shallower drawdown is mechanical: a trend filter sits in cash for a good share of the time, and anything that is invested less falls less. The comparison that separates timing skill from simply holding less is an exposure-matched benchmark. If the strategy averaged, say, 55% invested across the six sleeves, hold 55% SPY (or 55% of the equal-weight basket) and the rest in cash, and compare CAGR and max drawdown against that. If the rule still beats the scaled-down benchmark on drawdown at similar return, the timing is doing work.
Two smaller things that move the numbers over 2023-2026: whether idle cash earned T-bill interest in the backtest (4-5% a year on the uninvested share is not small next to a 4-point CAGR gap), and that a three-year max drawdown is usually one episode, so the halving rests on how the rule handled a single selloff. Showing the drawdown in that episode next to the exits it made would make the claim easy to check.
Fully agree, the strategy is too often in cash. Which is why I put a hypothetical scenario of the returns - i.e. when we are in cash, we invest in bonds or some other 'safe' liquid asset. But I like your idea, if I understood it correctly, we compare the strategy to the benchmark by exiting the benchmark when the strategy sits in cash - then compare the performance results?
Yes, that's it. Build a second benchmark that holds SPY on exactly the days the strategy is invested and sits in cash (or your bond proxy, same as the strategy) on the days it's out. Both lines then have identical exposure, so the only difference left is which sectors you held while invested. If the rotation still beats that exposure-matched SPY on return per unit of drawdown, the sector pick is adding something; if they come out about even, the drawdown benefit is the cash, and a plain "SPY with the same exit days" would do the same job.
To test the exit timing itself, keep the same number of cash days but place them at random, a few hundred times, and see where your real max drawdown falls in that spread. If it sits well below most of the random versions, your exits are landing before the falls rather than just reducing exposure.