The overlay that did its job, and the half of the record nobody shows you
A simple in-or-out trend rule on a broad US stock index cut the worst fall in half over 33 years. All of its advantage over holding the index came before 2009.
By Baris Demirci · · Rev00
About the figures in this article
Every figure and chart below is a simulated backtest on one exchange-traded index fund, not the performance of a live account. Backtests are run with hindsight, cannot fully capture costs or fills, and are not actual trading. The limits are listed at the end. Read them.
Nothing here is a prediction, a promise, or an indication of future results. Trading involves risk, including total loss of capital. This material is educational only. StratoQuant is not a licensed investment adviser, does not provide personalised investment advice, and does not manage money or access client accounts. See the full disclaimer.
The idea
Some systems don’t pick stocks at all. They hold one broad stock index and make a single decision: be in the market, or be out of it in cash. A trend rule decides which.
This kind of overlay is often recommended as a base layer under everything else you own, because long backtests show it cutting the big falls. It decides whether all of your money is in the market at once, so it deserves a high bar of evidence. We tested it ourselves.
How we tested it
- One US large-company index fund, dividends included, from January 1993 to September 2026 (33.6 years).
- The system is either fully in or fully out. No shorting, no leverage.
- It reads the trend at the close and acts at the next day’s open, so it never uses information it would not have had.
- Commissions and slippage are charged on every trade. Cash earns a short-term government rate.
- The main setting is the one from the published research, fixed before the test. About 300 variations were run around it to see how sensitive it is.
What we found
1. It halves the damage, at a small cost
Over the full period the system grew 9.4% a year, against 10.8% for simply holding the index. In exchange, its worst fall was 24% instead of 55%, and its longest stretch below a previous high was 3.2 years instead of 6.6. It gave up about 1.4 points a year and roughly halved every measure of downside. On that, it does what it is known for.
2. All of the advantage came before 2009
| Period | System, per year | Index, per year | System worst fall | Index worst fall |
|---|---|---|---|---|
| 1993 to 2008 | 10.4% | 6.5% | −19.0% | −50.8% |
| 2009 to 2026 | 8.5% | 14.7% | −24.4% | −33.7% |
| Full period | 9.4% | 10.8% | −24.4% | −55.2% |
Before 2009 the system beat holding the index by about 4 points a year, with a third of the drawdown. Since 2009 it has trailed by about 6 points a year, and most of its drawdown advantage has gone. In the 17 full years since 2009 it came out ahead in only two, each time by a fraction of a point. The 33-year average hides a complete reversal.
3. A few years decide everything
Two calendar years supply almost the whole risk advantage. The rule does well in long, steady declines and badly in sharp falls that bounce back quickly. In the most recent big drop it switched in and out six times and finished that year behind the index.
4. The popular tweaks didn’t help
- Checking every day instead of on a fixed schedule: lower growth, deeper falls, and about 4.6 times as many trades.
- Ignoring small moves with a “neutral zone”: this recreated the fixed-schedule version almost to the decimal. Both ideas do the same job, so using both adds nothing.
- Holding bonds instead of cash while out: slightly more growth, but a worst fall about 15 points deeper. In the last big drop, bonds fell at the same time as stocks.
- The best-looking settings in the whole study traded fewer than a dozen times in 33 years and missed the one decline this rule is famous for. We discard them.
One detail did matter: counting interest on idle cash was worth about half a point a year. Many published comparisons leave it out.
What to take from it
- Decide first whether you want less risk or more growth. Here the answers point in opposite directions.
- Split your test at the point the idea became popular, and look at both halves.
- When a tweak looks better, check whether it simply trades less.
- If many nearby settings give identical results, the system has stopped doing anything.
- A safe-looking holding is still a position. Test it against the same falls.
Limits
This part matters most, so it is blunt.
- One index fund, one market, one currency. This is a single case study, not a general result.
- It is really about three events, not 33 years. Two long declines and one sharp one drive every number that matters.
- Only 23 trades in 33 years. Anything measured on 23 outcomes is uncertain, and no confidence intervals are given here.
- Simulated results only. No live account, no forward test, no capacity or tax analysis.
- The 2009 split was chosen after seeing the data. It describes this sample. It is not a rule for splitting future ones.
- Costs are modelled, not measured. Real fills at the open could be worse, and a real account would earn less interest on cash.
- About 300 variations were examined. The main setting was fixed in advance, which helps, but the others still shaped how we read the results.
- The century-long histories behind this idea’s reputation cannot be checked here. The fund only exists from 1993.
Written for people building their own systematic portfolio management process. Nothing here is a recommendation to buy or sell anything. See the disclaimer.