Sell in May and Go Away: does the six-month rule still work?

"Sell in May and go away" is the best-documented seasonal effect in equities, and the one most likely to lose you money if you follow it literally.

Both of those things are true at once, and the gap between them is the whole subject.

The claim

The rule splits the year into two six-month halves: November through April, and May through October. The claim is that the winter half carries almost all of the market's return, and the summer half carries almost none of it.

The full name is older and blunter: sell in May and go away, come back on St Leger's Day — a horse race run in mid-September, which tells you the rule started as a description of when London bankers took their holidays.

What the two halves actually returned

The numbers below are computed from the S&P 500 (^GSPC) daily closes, January 1990 to August 2026, on total price return excluding dividends. Every figure is reproducible from that series.

The winter half returned 2.4× the summer half on average across 36 years. That is the effect, and it is large.

Now look at the second row. The summer half's median is +3.53%, not a negative number — so in a typical year, sitting out May to October costs you money rather than saving it. The average gap is real; the instruction people take from it is not.

The part that holds up

The seasonal gap is real and it is old. It shows up across many decades, across most developed markets, and it survives being sliced by sub-period. That is unusual: most seasonal effects disappear the moment you check them outside the window they were discovered in.

It is also stable in the direction nobody expects. The effect has not vanished since being published, which is what normally happens to an anomaly once it is widely known.

The part that does not

Three problems, in order of how much money they cost:

The summer half is positive. 72% of summers since 1990 were up, averaging +2.71% with a median of +3.53%. The rule is usually stated as though May-to-October loses money. It does not — it just earns much less than the winter half. "Underperforms" and "loses" are very different instructions. Sitting in cash for six months to avoid a small positive return is a guaranteed cost against an uncertain benefit.

The variance is enormous. The best summer in the sample returned +22.1%, the worst -31.3%. A spread of 53 percentage points around an average of +2.71% means the average tells you almost nothing about any individual year. Any single year is nearly uninformative about the average, and you only get to live through single years.

Costs and taxes are not free. Two round trips a year, in a taxable account, against a gap of a few percentage points. Do that arithmetic before doing anything else, because it frequently kills the strategy on its own.

The better use of it

The rule is more useful as a weighting than a switch. Knowing that the summer months have historically delivered less return per unit of risk is a reason to size differently, or to be more selective about entries, not a reason to hold cash from May Day to Halloween.

It is also worth remembering that "the market" is not what most people own. The seasonal split varies a lot by sector — and the sector composition of an index changes over decades, which means part of what looks like a stable seasonal effect is really a changing mix underneath.

Check the six-month split on what you actually hold

1. 1Pick your ticker in the symbol selector.
2. 2Set the date range to at least 20 years, so you get 20 summers rather than a handful.
3. 3In Trading Analysis, set the period to May 1 – Oct 31 and note the win rate, median and worst year.
4. 4Change the period to Nov 1 – Apr 30 and read the same three numbers.
5. 5Compare the two medians, not the two averages. The average is where this rule gets oversold.

If the summer median comes out positive, as it does for the index, then sitting out costs money in the typical year and the rule is a weighting rather than a switch.

Open TradeSeasons

How to check it yourself

Take the instrument you actually hold, not the index. Compute the return from the start of May to the end of October in each of the last twenty or thirty years. Then look at three numbers together: the median, the win rate, and the worst year. If the median is small and the worst year is large, you have found a statistic, not a strategy.

That test is worth running on any seasonal claim before you act on it — here is the method in full.