Does "sell in May and go away" work? Half of it does. Since 1993, $SPY's November–April half returned an average of 7.28% on a total-return basis, against 4.32% for May–October. But the "go away" part is expensive. Sitting in cash at 0% every May–October turned $10,000 into $89,181 by the October 30, 2026 close, while simply holding $SPY turned the same $10,000 into $312,040.
The weak half is weaker, but it is not a losing half. That one distinction explains almost everything below.
Where the "Best Six Months" idea comes from
"Sell in May and go away" is an old Wall Street adage. The tradable version comes from the Stock Trader's Almanac. Its publisher says on its strategy page that it first discovered in 1986 that most market gains occur from November through April. The rule it describes is to hold the Dow Jones Industrial Average from November 1 to April 30 and switch into fixed income for the other six months. A StockCharts article credits the discovery of the "Best Six Months" switching strategy to Yale Hirsch of the Almanac in its 2016 backtest. That test also found that the switch had a lower compound return than buy-and-hold but a smaller maximum drawdown.
This piece rebuilds the test from raw ETF data for four US index trackers: $SPY (S&P 500), $QQQ (Nasdaq-100), $IWM (Russell 2000) and $DIA (Dow).
How the halves were measured
- Data: Yahoo Finance historical data, daily bars through the November 18, 2026 close. $SPY starts January 1993, $DIA January 1998, $QQQ March 1999 and $IWM May 2000.
- Prices: Adjusted close (dividends and splits folded in), so every half-year return below is a total-return proxy unless labeled "price".
- May–October: the last close of April to the last close of October.
- November–April: the last close of October to the last close of the following April.
- Only complete halves count. That gives $SPY 33 November–April halves (1993–94 through 2025–26) and 34 May–October halves (1993 through 2026).
Nov–Apr vs May–Oct: the scoreboard
| ETF / half | Halves | Average | Median | % positive | Worst |
|---|---|---|---|---|---|
| $SPY Nov–Apr | 33 | 7.28% | 8.18% | 78.8% | −10.89% (2000–01) |
| $SPY May–Oct | 34 | 4.32% | 4.94% | 79.4% | −29.21% (2008) |
| $QQQ Nov–Apr | 27 | 6.33% | 8.00% | 70.4% | −43.52% (2000–01) |
| $QQQ May–Oct | 28 | 6.42% | 7.86% | 75.0% | −30.24% (2008) |
| $IWM Nov–Apr | 26 | 7.04% | 6.12% | 65.4% | −18.49% (2021–22) |
| $IWM May–Oct | 26 | 2.97% | 1.48% | 61.5% | −26.01% (2002) |
| $DIA Nov–Apr | 28 | 6.63% | 5.31% | 78.6% | −11.04% (2008–09) |
| $DIA May–Oct | 29 | 2.74% | 3.11% | 75.9% | −26.10% (2008) |
Three things jump out.
The gap is real on three of four. $SPY, $IWM and $DIA all show November–April averages roughly 3 to 4 percentage points above May–October. $IWM's median May–October half was just 1.48%.
The weak half still usually goes up. $SPY's May–October half was positive 79.4% of the time, slightly more often than its November–April half at 78.8%. The seasonal edge is about size, not direction.
$QQQ breaks the pattern. Its two halves average 6.33% and 6.42%, essentially a tie, and its worst half of all was a November–April: −43.52% from October 2000 to April 2001, during the dot-com bust.
Head to head, pairing each November–April with the May–October that follows it (one November-to-October cycle), November–April won 23 of 33 cycles for $SPY, 15 of 27 for $QQQ, 17 of 26 for $IWM and 19 of 28 for $DIA. The pairing matters, though. Matching each November–April with the May–October that came before it, $SPY's count drops to 19 of 33.
The switch vs buy-and-hold: growth of $10,000
The test: hold the ETF from the last October close to the last April close, then sit in cash earning 0% until the next October close. Buy-and-hold runs over the same dates. Both use month-end adjusted closes, and drawdown is measured on month-end values.
| ETF (period) | Switch: Nov–Apr only | Buy-and-hold | CAGR switch / hold | Max drawdown switch / hold |
|---|---|---|---|---|
| $SPY (Apr 1993–Oct 2026) | $89,181 | $312,040 | 6.75% / 10.82% | −30.48% / −50.78% |
| $QQQ (Apr 1999–Oct 2026) | $37,091 | $152,702 | 4.88% / 10.42% | −59.27% / −81.08% |
| $IWM (Apr 2001–Oct 2026) | $46,807 | $80,154 | 6.24% / 8.50% | −36.03% / −52.47% |
| $DIA (Apr 1998–Oct 2026) | $54,059 | $103,086 | 6.10% / 8.53% | −28.33% / −47.03% |
The switch lost on ending wealth in all four ETFs. It won on drawdown in all four.
Here is why it lost. Buy-and-hold is simply the two halves multiplied together. For $SPY, the November–April-only account grew 8.92 times and a May–October-only account grew 3.50 times ($34,990 from $10,000). Then 8.92 × 3.50 = 31.2, which matches buy-and-hold's 31.2 times ($312,040). Skipping May–October did not dodge a loss. It gave up a 3.5-times multiplier.
On $QQQ, the May–October-only account ($41,170) actually beat the November–April-only account ($37,091). For drawdown context, see the drawdown recovery math: cutting a −50.78% hole to −30.48% means needing roughly +44% to recover instead of +103%.
Every result here assumes no costs, taxes or slippage, trades at the month-end close, cash at 0% (a real T-bill yield would have narrowed the gap), surviving ETFs and a single data source.
Where 2026–27 stands so far (partial)
The "worst six months" of 2026 were not bad. $SPY returned +5.69% from April 30 to October 30, 2026 on an adjusted-close basis (+5.16% on price, from $718.66 to $755.72). That is a reminder of why sitting it out costs money.
The 2026–27 November–April window opened at the October 30, 2026 close. Through the November 18, 2026 close, it stands at −3.38% for $SPY (755.72 to 730.20), −4.49% for $QQQ, −3.30% for $IWM and −2.98% for $DIA, all price-only. That covers less than three weeks of a six-month window. It is a partial reading, not a forecast, and in the record above the full $SPY November–April half finished negative in 7 of 33 halves.
Bull/Bear Verdict
Bull Case: Where it helps: as a risk-and-sizing lens, the seasonal split holds up. November–April out-averaged May–October by roughly 3 to 4 points on $SPY, $IWM and $DIA, and the switch cut $SPY's month-end maximum drawdown from −50.78% to −30.48%. Traders use it to lean harder into setups during the stronger half.
Bear Case: Where it misleads: "go away" implies May–October loses money, and historically it does not. $SPY's weak half was positive 79.4% of the time, so exiting it turned a $312,040 result into $89,181. On $QQQ the seasonal edge vanished entirely.
FAQ
Does "sell in May and go away" actually work?
As a description, mostly yes. As a strategy with 0% cash, no. From 1993 to October 2026, $SPY's November–April half averaged 7.28% against 4.32% for May–October. But holding only November–April compounded at 6.75% a year, against 10.82% for buy-and-hold.
What was the worst May–October for $SPY?
2008. $SPY lost 29.21% on an adjusted-close basis from April 30 to October 31, 2008, a stretch at the heart of the 2008 financial crisis.
Who came up with the Best Six Months strategy?
The Stock Trader's Almanac says it first discovered the November–April pattern in 1986, and StockCharts credits Yale Hirsch of the Almanac. The Almanac's rule holds the Dow from November 1 to April 30 and fixed income for the rest of the year.
Is May–October negative on average?
No. $SPY's May–October half averaged +4.32% and was positive in 27 of 34 years since 1993. $DIA's was positive 75.9% of the time and $IWM's 61.5%. The half is weaker, not negative, which is why sitting in cash through it has historically cost more than it saved.
Does the pattern work on small caps and the Nasdaq-100?
On $IWM the gap was the widest: a 7.04% average for November–April against 2.97% for May–October. On $QQQ it was essentially a tie, at 6.33% against 6.42%. More seasonal and trend studies are in the Markets section and on the Traders On Trend homepage.