Quantum BiasSell in May and Go Away: Does It Work on the Nasdaq-100?
In this article
Not on the Nasdaq-100. From January 2000 to September 2026, the May–October months actually returned a bit more than November–April. Switching to Treasury bills every summer would have cut the annual return roughly in half.
| Question | Data | Answer |
|---|---|---|
| Do Nasdaq-100 stocks earn more from November to April? | QQQ monthly total returns, Jan 2000 – Sep 2026, 321 months | No: +0.85% vs +1.01% a month, t = −0.2 |
The study we tested
Sven Bouman and Ben Jacobsen published The Halloween Indicator, "Sell in May and Go Away": Another Puzzle in the American Economic Review (vol. 92, no. 5, 2002). Using MSCI indexes from 1970 to 1998, they found the effect in 36 of 37 markets. In the U.S., November–April returns ran 17.1% a year against 6.0% for May–October.
The Nasdaq-100 barely existed as something you could trade in their sample. So we asked: does the rule hold on the index NQ traders follow, after the study came out?
What we found
- November–April: 160 months, average +0.85% a month.
- May–October: 161 months, average +1.01% a month.
- Difference: −0.16 points a month, t-statistic −0.22. No effect, and the sign is flipped.
- Season by season: winter beat summer in 13 of 25 full seasons (November 2000 to October 2025). A coin flip.
- By period: 2000–2012, winter +0.29% vs summer −0.03% a month (t = 0.26). 2013–2026, winter +1.37% vs summer +1.99% (t = −0.77).
Against every other six-month window
The fair test isn't "winter vs. summer". It's "Nov–Apr vs. any other six months". We held QQQ for each of the 12 possible six-month windows and Treasury bills the rest of the year.
Exact values from table 2 in this article. Bar lengths share a zero baseline; no new calculation or simulated price path. Read the table and its stated sample and limits in this article.
Nov–Apr finished 9th of 12. No window beat just holding QQQ all year, because every window spends half the year out of the market.
What the trading rule would have earned
| Annual return | Volatility | Worst drawdown | |
|---|---|---|---|
| Buy and hold QQQ | 8.8% | 22.9% | −81.1% |
| Nov–Apr in QQQ, May–Oct in T-bills, before costs | 4.8% | 16.5% | −57.2% |
| Same, 0.10% cost per switch | 4.6% | — | — |
The rule did lower risk: less volatility and a smaller worst loss, partly because it was in T-bills during May–October 2002 and May–October 2008, when QQQ fell 22.6% and 30.2%. It paid for that with about four points of return a year.
What this means if you trade NQ
A seasonal pattern found on broad markets doesn't automatically carry over to a concentrated tech index. Seasonality can still be useful as context, but this one hasn't earned its reputation on the Nasdaq-100 since 2000.
Limits
- QQQ, not NQ. Same index; the futures contract would also carry financing and roll costs.
- Monthly data. Dividends are added in the month of their ex-date, which is an approximation.
- No taxes. Two switches a year would trigger taxable events in most accounts.
- One start date. Our sample starts at the peak of the dot-com bubble, which weighs on early winters and summers alike.
Method and sources
- QQQ monthly prices: Alpha Vantage
TIME_SERIES_MONTHLY, adjusted by us for the March 20, 2000 two-for-one split; dividends from Alpha VantageDIVIDENDS. - Treasury bills: 3-month yield, monthly average, Alpha Vantage
TREASURY_YIELD. - Original study: Bouman and Jacobsen, The Halloween Indicator, "Sell in May and Go Away": Another Puzzle, American Economic Review, 2002.
Research and education only. Not financial advice. Past results do not predict future returns. A backtest is not real performance.
The week just closed, summarized, and the economic calendar for the week ahead.