Should you sell when the market falls below its 200-day average?

What it has done since it got popular

We ran it on dividend-adjusted SPY prices from 2017. The rule returned 8.54% a year. Holding the index returned 15.01%. The rule gave up around six and a half points a year against doing nothing.

What it bought in exchange was slight: a worst fall of 21.9% against 23.9% for holding.

There were ten complete round trips out and back. Nine bought back in at a higher price than they had sold at. The tenth was 2022, seven months in cash, and it finished flat, down 0.2%.

2022 was the only genuine bear market in the sample. The rule lost more than holding did: down 20.1% against the index’s 18.2%. It was invested for the falls of January, February, April and December, and sitting in cash for the rises of March, July, October and November.

2020 is the year people remember it for. The rule sold at the end of February, missed March’s 12.5% fall, then also missed April’s 12.7% and May’s 4.8% rises, and bought back 3.3% above where it had sold. It dodged the crash and paid for the privilege. The daily version crossed the line five times in eight days that March.

What the rule’s author actually claimed

His own table, 1972 to 2005 across five asset classes, reports returns approximately the same as buying and holding, with lower risk. Not higher returns. Ten years on he noted the rule "went on to underperform stocks six of the next eight years", and wrote that had it done badly, "likely no one would be reading this article". On tax he gives no figure, and tells individuals to run it inside a retirement account.

Is there a real effect underneath?

Possibly, and here the comparison you pick decides the answer. Because the rule sits in cash about a quarter of the time, it holds less in shares on average than someone who never sells. Over the century of US data it lost 0.73 points a year against simply holding, and gained 0.74 points a year against a fixed blend that holds the same average amount in shares. Both are true. Only the second is a claim about the rule rather than about owning shares.

Even that second number settles nothing. The smallest gap the test could reliably have seen was 3.03 points a year, because a century of monthly data contains only 73 actual decisions. We could not tell, which is different from finding nothing there.

The one study with enough history to settle it pools 16 countries across 148 years, and it does find something: 0.97 points a year better than staying invested, and it holds up. Three caveats travel with it. Annual data cannot carry a ten-month average, so the rule tested there is a simpler one-year version. The underlying data has no fees, spreads or taxes in it and some of it was reconstructed from newspapers. And at annual spacing the rule had a deeper worst fall than its comparison in 9 of the 16 countries.

Pushing the other way: when a researcher rebuilt a well-known favourable study with a look-ahead mistake removed, the advantage disappeared and the extra return turned negative. A separate search over 7,846 trading rules on a century of daily prices found its winner had about a 90% chance of being luck once you count how many rules were tried. Its best rules used windows of two to five days. The 200-day average is folklore that became everyone’s starting point, not a length any search selected.

Why it still fails a taxable investor

Every exit in a taxable account realises gains you would not otherwise have owed tax on. We put that cost at 1.92 points a year, twice the size of the 0.97-point effect the best-powered study found. Over the long record, 58 of 73 exits lost money.

What it genuinely buys

Across the long US history the rule’s worst fall was 43.1%. The fixed blend holding the same average amount in shares fell 71.6%, and holding throughout fell 83.7%. That is mechanical rather than a matter of prediction, which is why it holds up when the return does not, and it repeats in every market tested at monthly spacing.

There is a cheaper way to buy most of that. Simply holding a smaller fixed share of stocks, with no rule, no signal and no trading, took the worst fall to 39.1% over the four decades to 2026. The timing rule went fifteen points deeper still, and charged 1.2 to 1.9 points a year in tax for the difference.

A related rule, dual momentum, has returned 9.17% a year since publication, against 14.87% for the index and 9.65% for a plain 60/40 mix, at about the same worst fall as the 60/40.

The strongest honest version of the case for timing is this: I will probably earn less than holding, I will trade about once a year and be wrong most of those times, I should only do it where tax cannot reach me, and what I am buying is a smoother ride and a rule that stops me panicking. Anything stronger is contradicted by the authors’ own tables.

What would change our mind

A test with enough history to see an effect smaller than 3 points a year, run on data the rule was not designed against, showing an advantage that survives tax. Or evidence that the long international result holds up in the decades since it was published rather than only before.

Where these numbers come from

• Our own replication on dividend-adjusted SPY daily prices, 2017 to 2026, and on Kenneth French’s US market series back to 1926.
• Meb Faber, "A Quantitative Approach to Tactical Asset Allocation" (2007) and his ten-year review of it.
• The Jorda, Schularick and Taylor macro-history database, 16 countries, 1870 to 2020.
• The corrected replication of a widely cited favourable moving-average study.
• The survey testing 7,846 technical trading rules on a century of daily US prices.
• Gary Antonacci’s published dual momentum rule, run forward from its publication date.