Skip to main content
Seasonality Investing

Seasonality Investing · Stock Seasonality, 8 lessons · Lesson 2

Sell in May was real. And then it cooled

Nov–Apr and May–Oct were clearly different over 76 years. But over the last 10 years, the difference disappeared. Read it not as "sell" but as "expected returns were lower."

15 min readLesson 2 of 8As of: S&P 500 1950–2025, 76 periods · KOSPI 1997–2025, 29 periods · calculated 2026-10-09✓ Read

QToday's question

Should you really follow "Sell in May and go away"?

$10,000 became $1.17 million on one side and $34,000 on the other. Same S&P 500, same 76 years; the only difference was which half of the year you held. This lesson starts from this single number, then follows the evidence that the gap was real and the reversal that cooled it over the last 10 years.

MAPLearning map for this lesson

We build it up in three steps. At the end of each step, a 'Quick check' lets you test your understanding before moving on.

Before you start

In Lesson 1, Why is September so disliked?, we said that seasonality is not a prophecy but a tilt in probabilities, and that it is read through four numbers: average, , and . This lesson applies those four numbers to half-years instead of single months. September, the weakest month, falls within May–Oct, while the strong months of November, December and April fall within Nov–Apr.

  1. 1 · One number$1.17M : $34K

    You can recall how much $10,000 differed depending on which half-year you picked, and state the conditions behind that number (price index, zero interest on cash).

  2. 2 · Insight · 3 · Data36 of 37 countries

    You know what compares with what, and can read the results of 76 S&P 500 periods and 29 KOSPI periods split by era.

  3. 4 · Reversal · 5 · PracticeIt was different, but the gap narrowed

    You can confirm that the gap disappeared over the last 10 years, and compare fixed-date trading with -confirmed trading yourself.

Build knowledge9 concepts built so far · 4 added in this lesson

Seasonality = a tilt in probabilitiesAverage, median, win rate, sample sizeComparing against a baselineSince 1950 vs the last 20 yearsNot "sell" but "lower your expectations"Nov–Apr vs May–Oct half-year comparisonNEWHalloween effect (36 of 37 countries)NEWThe gap vanishing over the last 10 yearsNEWFixed-date vs MACD-confirmed switchingNEW

1 · One number$1.17M : $34K

$1.17M : $34KResult of investing $10,000 only in Nov–Apr vs only in May–Oct (USD)S&P 500, 76 periods, 1950–2025 · price index basis, excluding dividends and costs, zero interest on cash in the off half-year

Suppose you put $10,000 into the S&P 500 in 1950. If you held it only in Nov–Apr each year and rested in cash during May–Oct, it would have grown to about $1.17 million by the end of 2025. Conversely, if you held it only in May–Oct and rested during Nov–Apr, you would have had about $34,000. Same index, same 76 years; the only difference was which half of the year you chose.

These numbers come with conditions. They use a price index that excludes dividends, trading costs and taxes, and assume zero interest during the off half-year. Also keep in mind that multiplying the two results gives the buy-and-hold result (about 400x). Even holding only May–Oct, the money grew 3.4 times.

Quick checkFrom the sentence "$10,000 became $34,000 by investing only in May–Oct," can we say that May–Oct was a period when you lost money?
No. The money grew 3.4 times over 76 years, so May–Oct was positive too. It just grew much more slowly than Nov–Apr.

2 · InsightThe same in 36 of 37 countries

is the observation that, when you split the year into two half-years, Nov–Apr and May–Oct, returns in the first half-year were higher than in the second. It starts around October 31, so it is also called the "Halloween effect," and the Stock Trader's Almanac named Nov–Apr the "." The key point is that the benchmark is not '0' but the opposite half-year. If you miss this, the saying gets passed along as "always sell in May."

Easy analogy · A restaurant's busy and slow seasons

A seaside restaurant has its peak season in summer. That doesn't mean closing in winter is a good idea; customers still come in winter, and sales are just 'a bit lower.' May–Oct was the slow season, not a season when business doesn't work.

This gap is not just a US story. Bouman & Jacobsen (2002) reported that in 36 of 37 countries' stock indexes, Nov–Apr returns were higher than May–Oct, and that the same phenomenon has been observed in the UK since 1694. Zhang & Jacobsen (2021) analyzed 62,962 observations across more countries and a longer period and found that Nov–Apr returns were about 4 percentage points higher on average. It is too wide and too long-lived to call a pattern found by chance after digging through one or two markets.

However, the cause has not yet been established. Explanations such as reduced trading and attention during summer vacations, and inflows of year-end bonuses and pension money are mentioned as candidates, but all of them are hypotheses. So this is a pattern with strong evidence but an unclear cause, and it is better used as a filter for other signals than as a standalone trading signal.

What Sell in May saysThe average return in Nov–Apr was higher than in May–Oct · It does not mean that 'May–Oct lost money'
Quick checkThe fact that Sell in May has been observed across many countries over hundreds of years raises what, and still doesn't tell us what?
It lowers the odds that it's a coincidence, but it doesn't tell us why it happens. Since the cause isn't established, it's hard to know in advance when it will weaken. That's why we use it as a filter, not a standalone signal.

3 · Data76 pairs of half-years

Splitting the S&P 500 into 76 half-years from 1950 to 2025, Nov–Apr averaged +6.95% with a median of +6.18% and rose in 58 of 76 periods (win rate 76%). May–Oct averaged +2.08% with a median of +2.68% and rose in 50 periods (win rate 66%). The average and median point in the same direction, and the sample of 76 is ample. Still, even a 66% win rate for May–Oct is by no means low.

How $10,000 (in dollars) grew · Nov–Apr only vs May–Oct only vs buy and hold

Hold Nov–Apr onlyHold May–Oct onlyBuy and hold

Calculated directly from month-end values of Yahoo Finance ^GSPC and ^KS11 daily closes (price index, excluding dividends, costs and taxes, zero interest on cash in periods not held). The vertical axis is a log scale. Retrieved 2026-10-09.

Splitting the KOSPI the same way from 1997 to 2025, Nov–Apr averaged +12.54% (up in 20 of 29 periods, 69%) and May–Oct averaged −0.15% (up in 12, 41%). The May–Oct median was −3.89%, so in Korea this half-year really did tilt toward the negative side. One interpretation is that Korea-specific schedules, such as year-end large-shareholder capital gains tax selling and re-buying at the start of the year, overlap, but this is a hypothesis, and Lesson 6 covers it in detail. Also note that the sample is only 29 periods, and one or two big moves, such as the currency crisis and the financial crisis, shake the numbers.

When you split the periods, the US picture changes.

S&P 500 periodNov–Apr avg (win rate)May–Oct avg (win rate)Gap
1950–1989 (40 periods)+7.12% (75%)+1.27% (60%)+5.85%p
1990–2005 (16 periods)+7.77% (88%)+1.92% (69%)+5.85%p
2006–2025 (20 periods)+5.96% (70%)+3.85% (75%)+2.11%p
2016–2025 (10 periods)+6.78% (70%)+6.93% (90%)−0.15%p

S&P 500 price index, simple average of half-year returns. The year refers to the year containing May–Oct. Calculated directly; retrieved 2026-10-09.

Quick checkThe KOSPI's gap (+12.7pp) is much larger than the S&P 500's (+4.87pp). Does that mean Korea has the more reliable pattern?
A big difference is not the same as a reliable one. The KOSPI sample is only 29 periods, so a single year like 1998 (Nov–Apr +95.6%) shakes the average a lot. Look at the median and the results by period together.

4 · ReversalOver the last 10 years, the gap disappeared

The reversal. Looking at only the 10 periods from 2016 to 2025, the S&P 500 averaged +6.78% in Nov–Apr and +6.93% in May–Oct, so May–Oct was actually slightly higher. In 2024 (May–Oct +13.30%, followed by Nov–Apr −2.39%) and 2025 (+22.83% vs +5.39%), May–Oct led by a wide margin two years in a row.

Looking only at the 76-year average, Sell in May looks solid, but when you split the periods, the gap shrank from +5.85 percentage points in 1950–1989 to +2.11 percentage points in 2006–2025, and disappeared over the last 10 years. A widely known pattern may have weakened as more people traded on it, or the small sample of the last 10 years may be a coincidence. We can't say for sure which it is now. Interestingly, the KOSPI still showed a gap in 2016–2025, with Nov–Apr at +12.90% and May–Oct at +3.29%, but with only 10 periods, it is hard to put much weight on it.

Each year's two half-years and the 10-year average gap

Nov–Apr (blue) − May–Oct return difference10-year average gap (moving average)
Each bar is one year. A long bar upward means Nov–Apr was better than May–Oct that year, and a long bar downward means May–Oct was better. Tap a bar to see it.

The year refers to the year containing May–Oct. Example: the 2024 bar = the difference between May–Oct 2024 and Nov 2024–Apr 2025. Calculated directly from the same data.

Common misconceptions

"Can selling in May help you avoid losses?" The S&P 500's May–Oct averaged +2.08% with a 66% win rate. More often, selling meant sidestepping gains rather than losses.

"It has been right for hundreds of years, so it will be right this year too?" Seasonality is not a forecast but a tilt in probabilities. There are stretches, like the last 10 years, when the tilt disappears.

"Korea's difference is bigger, so it's more certain?" A large difference is not the same as a reliable one. The KOSPI sample is 29 periods.

Quick checkThe gap changed from +5.85pp to −0.15pp. Which is right: "the 6-month cycle is over" or "it still holds"?
It's hard to claim either. Take the possibility that it has weakened seriously, but the last 10 periods are a small sample. Since the tilt has shrunk, it is safer to use it as a lighter-weight filter.

5 · One thing to practiceFixed date vs MACD confirmation

Today's practice has two parts. First, read "sell in May" as "expected returns were lower in May–Oct." Then compare, yourself, buying and selling purely by date against buying and selling after confirming the trend.

The Stock Trader's Almanac's 6-month strategy holds stocks in Nov–Apr and moves to bonds or cash in May–Oct, but the actual trading days are set not by the date but by a signal. That is, once October arrives you wait for a MACD buy signal, and once April arrives you wait for a sell signal. The exercise below simplifies this idea and reproduces it using S&P 500 daily closes. Across 76 periods, the MACD approach averaged +7.10% (win rate 72%), similar to fixed dates (+6.95%, 76%), and cumulatively $10,000 became about $1.31 million versus $1.17 million. Change the period and see for yourself that the order flips, and that both approaches trail buy and hold in many stretches.

Exercise · Fixed-date approach vs MACD-confirmation approach

Start1950
End2025
Fixed date (enter 11/1 · exit 4/30)-
MACD confirmation (buy signal after October · sell signal after April)-
YearFixed dateMACDMACD entry/exit

A simple educational backtest. S&P 500 price index, excluding dividends, trading costs, taxes and cash interest; MACD is based on daily-close exponential moving averages (12 and 26 days) with a 9-day signal line. It differs from the actual Almanac strategy (Dow index, detailed rules).

SituationHow to use the 6-month cycle
Late October to April, trend also favorableKeep contributing and buying as planned. Treat seasonality as no more than one extra confirming signal.
Entering May–Oct, trend-weakening signalInstead of selling everything, adjust your weight by slowing new purchases or raising your stop-loss level.
May–Oct but the trend is strongPut trend ahead of seasonality. There were years, like 2024 and 2025, when May–Oct was better.
Accounts with heavy taxes and costsFrequent all-in switching easily loses its edge to trading costs and taxes, so avoid it.

This is an educational example and does not recommend any specific trade. Lesson 7 wraps this up into a three-line 'condition, action, invalidation' rule.

The figures in this lesson are on a price-index basis, so dividends are left out. Including dividends would raise returns in both half-years, and a switching strategy earns interest on the cash period but pays trading costs and taxes. Including costs and taxes reduces the switching strategy's edge.

Quick checkIf you choose the last 10 years (2016–2025), both the fixed-date and MACD approaches trail buy and hold by a wide margin. How should you use this result?
Use it as grounds for not treating the 6-month cycle as an all-in trading rule. In the last 10 years the May–Oct average was high too, so the strategy that sat out that stretch fell behind. Use it as a filter that adjusts your weight slightly, along with a condition that invalidates it when the trend points the other way.
One sentence to take from Lesson 2It was different, but the gap narrowed

✓What we learned today

  • If you had put $10,000 in only for Nov–Apr starting in 1950, you would have had about $1.17 million, and only for May–Oct, about $34,000 (price index, excluding dividends and costs, zero interest on cash).
  • What Sell in May compares is not '0' but the opposite half-year. The S&P 500's May–Oct also averaged +2.08% over 76 periods, with a 66% win rate.
  • Bouman & Jacobsen (2002) reported that Nov–Apr was about 4 percentage points higher in 36 of 37 countries, and Zhang & Jacobsen (2021) reported the same across 62,962 observations. The evidence is broad, but the cause is not established.
  • The reversal: the S&P 500's gap shrank from +5.85 percentage points in 1950–1989 to +2.11 percentage points in 2006–2025, and in 2016–2025, with Nov–Apr at +6.78% and May–Oct at +6.93%, the difference disappeared.
  • Practice: read "sell in May" as "expected returns were lower in May–Oct." Use it as a filter that adjusts your weight, not as an all-in switch.
Continues in the next lesson · Lesson 3Santa comes, and January speaks

At the heart of the strong Nov–Apr period is the turn of the year. In Lesson 3, we check in turn the 76% win rate of the 7 trading days of the Santa Claus rally, the January barometer (the idea that January tells you about the whole year), and the small-cap January effect that has already vanished.

Quiz5-question quiz

Pick an answer and it's graded right away with an explanation.

Q1. Which best describes what Sell in May (the 6-month cycle) actually says?

Q2. If you had held the S&P 500 only in May–Oct starting in 1950, about how much would $10,000 have become?

Q3. When reading the KOSPI's Nov–Apr average of +12.54% and May–Oct average of −0.15%, what should you look at first alongside them?

Q4. Looking at only the 10 periods from 2016 to 2025, how did the averages of the two half-years look for the S&P 500?

Q5. What is the most appropriate reason the Almanac's 6-month switching strategy also uses a MACD signal?

NextTerms · Next lesson · Related tools

Sell in May (Halloween effect)Sell in May, Halloween indicator

A six-month seasonal pattern in which returns in November–April were higher than in May–October. It has been observed in many countries, but its cause has not been established.

Best Six MonthsBest Six Months

The name the Stock Trader's Almanac gives to November through April. Since 1950, the S&P 500's average return in November–April has been higher than in May–October, but May–October has also averaged positive.

Seasonal switching strategyseasonal switching

A method of holding stocks in strong periods and bonds or cash in weak periods. It involves trading costs and taxes, and in years when stocks rise even during the weak period, you miss those gains.

MACD (moving average convergence divergence)Moving Average Convergence Divergence

An indicator that spots trend reversals using the difference between the short-term (12-day) and long-term (26-day) exponential moving averages, along with its 9-day moving average (the signal line). A MACD crossing above the signal line is read as a buy signal, and crossing below as a sell signal.

Probability of a rise (win rate)Win rate, % up

The share of the whole sample in which the return was positive. Look at it together with the average to check whether the average is being pulled by a few large numbers.

Sample sizeSample size (n)

The number of observations used to calculate a statistic. In monthly seasonality it means how many times that month has occurred; the smaller the number, the more likely the result is due to chance.

Out-of-sample testingOut-of-sample test

This means checking whether the same pattern also shows up in a period other than the one where it was found (especially a recent period) or in another market. It is the basic way to filter out data-mining bias.

BaselineBaseline

The performance of "any random time," against which you compare a particular period. For example, only by comparing with the average of all months for the S&P 500 (+0.75%) and its share of up months (about 60%) can you tell whether a particular month was really strong.

Put what you learned to use right away

Compound Interest Calculator→Return & CAGR Calculator→Lesson 1 practice table→

For educational purposes only; not investment advice. Past statistics do not guarantee future returns. The S&P 500 was calculated directly from month-end values of Yahoo Finance ^GSPC daily closes (1950–2026), and the KOSPI from ^KS11 daily closes (December 1996–2026). Nov–Apr runs from the closing price at the end of October of a given year to the closing price at the end of April of the next year, and May–Oct from the end of April to the end of October. Price index basis, excluding dividends, trading costs and taxes; retrieved 2026-10-09. The MACD comparison is a simple educational backtest.