Seasonality investing · Stock seasonality, 8 lessons · Lesson 3
Santa comes, and January speaks
The 7 Santa-rally trading days were positive 76% of the time, versus 58% for any random 7 trading days. We use numbers to sort out what remains and what has disappeared among the year-end effects.
QToday's question
The Santa rally missed three years in a row. Is the year-end effect over?
At year-end, articles about the 'Santa rally' pour out, and at the end of January, articles saying 'January speaks for the whole year.' This lesson starts with one number, looks at the insight and data, and ends with a twist and one practice you can do today.
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.
In Lesson 1, Why is September so disliked?, we saw that probabilities of rising must be compared with a baseline, and in Lesson 2, Sell in May was real. And then it cooled, we saw that widely known patterns weaken as they get traded. This lesson uses those two tools to measure three year-end patterns.
- 1 · One number76% vs. 58%
You can remember the probability of a rise over the 's 7 trading days alongside the for any random 7 trading days.
- 2 · Insight · 3 · DataWhat January says
You can tell apart what the and the each claim, and check them with numbers by period.
- 4 · Twist · 5 · PracticeThe more textbook the pattern, the sooner it weakens
You'll read about the Santa rally that missed three years in a row and the vanished small-cap effect, and calculate this year's year-end 7 trading days and January barometer yourself.
1 · One number76% vs. 58%
The last 5 trading days of December and the first 2 trading days of the following January, these 7 trading days, are called the window. Yale Hirsch defined it in the 1973 edition of the Stock Trader's Almanac. For the S&P 500, this window ended positive in 58 of 76 cases from 1950 to 2025 (76%). The average was +1.25%.
This number alone looks impressive, but as we saw in Lesson 1, a probability of rising is never read in isolation. The probability of a rise when you pick any random 7 trading days over the same period, that is, the , was 58%, with an average of +0.25%. The Santa rally's 'strength' is not 76% but the difference from the baseline, about 18 percentage points, or about 1.0 percentage point in average return.
Quick checkWhat should you add to the statement "The Santa rally win rate is 76%" to make it accurate?
2 · InsightDoes January speak for the whole year?
Three named patterns are packed into the turn of the year. They look similar, but each claims something different.
| Pattern | What it claims | First documented by |
|---|---|---|
| The average and win rate over the last 5 trading days of December and the first 2 of January were high | Yale Hirsch, 1973 edition of the Stock Trader's Almanac | |
| In years when January rose, the average for the other 11 months was also high | Popularized by the Stock Trader's Almanac | |
| Small-cap returns in January were higher than large-cap returns | First recorded by Sidney Wachtel in 1942 |
The Santa rally and the January effect describe the returns of that period itself, but the January barometer claims that January 'foretells' the following 11 months.
The most interesting one is the January barometer. From 1951 to 2025, the S&P 500's January was positive in 45 years and negative in 30. In the 45 years when January rose, the average for the other 11 months (end of January to end of December) was +12.07%, and 87% ended positive. In the 30 years when January fell, the average was +2.07%, with 60% rising. The average difference between the two groups is about 10 percentage points. The saying "January speaks for the whole year" has at least 75 years of numbers behind it.
Easy analogy · year-end cleaning and new-year grocery shopping
At year-end you tidy the house and throw out things you don't use, and in the new year you go shopping to refill the fridge. The stock market is similar. At year-end, investors sell losing stocks to cut taxes (), and funds tidy their portfolios for their year-end reports. In the new year, bonuses and pension contributions come in fresh. Because the times for throwing out and refilling are tied to the calendar, the explanation goes that patterns form around the turn of the year.
The candidate causes overlap in three: taxes (selling to realize losses), money flows (bonus and pension inflows), and settlement and reporting (funds' year-end tidying). Because the cause can be explained, the year-end patterns are usable candidates. But having a cause is no guarantee that the pattern still exists. That is checked separately with data.
Quick checkThe saying "If January rises sharply, it will be a good year" corresponds to which of the three patterns, and how does it differ from the other two?
3 · DataRe-measuring 76 years by period
We measure the three patterns one by one, using the baseline and period splits. We always put the full period and the last 20 years side by side.
Santa rally: the gap versus the baseline has narrowed
In 1950–1989, the Santa rally had an 82.5% probability of rising over 40 observations and an average of +1.72%, a probability 25 percentage points higher than the baseline (57.0%). In 1990–2005 it was 68.8% over 16 observations and +0.92% (baseline 56.5%), and in 2006–2025 it was 70% over 20 observations and +0.55% (baseline 60.9%). The gap versus the baseline shrank from 25 percentage points to 9. The line in the chart below is the probability of a rise over the previous 10 years.
76 Santa rallies and the 10-year rolling probability of a rise · S&P 500
Calculated directly from Yahoo Finance ^GSPC daily closes (price index, excluding dividends and costs). Santa rally = last 5 trading days of December + first 2 trading days of the following January. Baseline = all possible windows of 7 consecutive trading days in 1950–2025 (19,114). Retrieved 2026-10-09.
January barometer: clear over the full period, smaller in the last 20 years
Split into a 2×2 table, of the 45 years from 1951 to 2025 when January rose, in 39 the other 11 months were also positive. But of the 30 years when January fell, in 18 the other 11 months were also positive. Narrowing the period to 2006–2025 (20 observations), the other 11 months averaged +10.91% in the 11 years when January rose and +8.98% in the 9 years when it fell, a difference of just over 2 percentage points.
January barometer 2×2 · S&P 500
Calculated directly from Yahoo Finance ^GSPC month-end closes (price index). January = end of January vs. end of the previous December; the other 11 months = end of December vs. end of January. Retrieved 2026-10-09.
Small-cap January effect: close to zero at the index level
We calculated, year by year, the , the Russell 2000 small-cap index's January return minus the S&P 500's return in the same month. The 1988–2005 average over 18 observations was +0.09 percentage points (ahead in 7 of 18), and the 2006–2026 average over 21 observations was +0.02 percentage points (ahead in 9 of 21). The medians were −0.14 and −0.12 percentage points respectively, both negative.
Russell 2000 January excess return (vs. S&P 500)
Calculated directly from Yahoo Finance ^RUT and ^GSPC month-end closes (price index, excluding dividends). 39 Januaries, 1988–2026. Retrieved 2026-10-09.
Quick checkIn 2006–2025, the Santa rally was positive in 70% of 20 observations, and the baseline for the same period was 61%. How many percentage points is this pattern's strength?
4 · TwistThe more textbook the pattern, the sooner it weakened
Three negative years in a row is a striking record, but the sample is only 3. Widening to the 20 observations since 2006, 14 were positive. So 'it shrank to only slightly better than the baseline' fits the data better than 'it is over.'
Why do famous patterns weaken first? If investors who know the year-end average is high buy a little earlier, the gains get pulled forward to mid-December and the share left for the actual 7 trading days gets smaller. This is a hypothesis, but it points the same way as the 'weaker predictive power after publication' reported by McLean & Pontiff (2016). The more widely known a pattern is, enough to make it into textbooks like the small-cap January effect, the sooner it feels this pressure.
The January barometer faces the same question. Over the full 1951–2025 period, the gap between years when January rose and years when it fell clearly remains at about 10 percentage points, but looking only at the 20 observations since 2006, that gap shrinks to just over 2 percentage points. With 20 samples it is too early to conclude, but rather than taking 'January speaks for the whole year' at face value, it is safer to read it as 'the average was higher in years when January rose.'
Common misconceptions
"January fell, so I should reduce stocks this year?" Of the 30 years from 1951 to 2025 when January fell, in 18 the other 11 months were positive, and the average was +2.07%. It was just a lower expected return.
"The Santa rally missed three years in a row, so it's over now?" We don't judge from 3 samples. It is true, though, that the gap to the baseline has shrunk to 9 percentage points.
"Are small caps favored in January?" The Russell 2000's January excess return was close to zero in both periods since 1988, and the median was negative.
Quick checkIf you pick only the three years 2023–2025 and compute the Santa rally average, it is negative. How should you treat this result?
5 · One thing to practiceThis year's year-end 7 trading days and the January barometer
There are two practices in this lesson. Recalculate the Santa rally record yourself for any period you like, and prepare in advance a check card to use this year-end and January.
Practice 1 · Build a Santa rally scorecard
| Year | 7-trading-day return | vs. baseline average |
|---|
Calculated directly from Yahoo Finance ^GSPC daily closes (price index). Baselines: 1950–1989 average +0.24%, 57.0% up; 1990–2005 +0.25%, 56.5%; 2006–2025 +0.27%, 60.9%; full period +0.25%, 58.0%. If the period differs from these, the full-period baseline is used.
Press the last-10-years button and you get an average of +0.48% and a 70% probability of a rise. Put next to +1.72% and 82.5% for 1950–1989, you can see how much the weight of the same pattern has changed.
Practice 2 · This year's year-end check card
Inputs and check states are saved only in this browser. Comparison numbers: Santa rally baseline, 2006–2025 average +0.27%, 60.9% up. January barometer 1951–2025: 45 years January rose → other 11 months +12.07% (87%), 30 years it fell → +2.07% (60%); 2006–2025: 11 up years +10.91%, 9 down years +8.98%.
The figures in this lesson are on a price-index basis, so dividends, trading costs and taxes are excluded. The Santa rally and the January barometer are observed only once a year, so even 76 years gives only 75–76 samples, and the last 20 years only 20. This is an educational example and does not recommend any particular trade.
Quick checkThis year's January was −3%. What should you write on the check card?
✓What we learned today
- The S&P 500's Santa rally 7 trading days had a 76% probability of rising over 76 observations in 1950–2025, with an average of +1.25%. The baseline for any random 7 trading days was 58% and +0.25%.
- January barometer: in the 45 years from 1951 to 2025 when January rose, the other 11 months averaged +12.07% (87% up), and in the 30 years it fell, +2.07% (60%). In 2006–2025, though, it was +10.91% versus +8.98%, a gap that shrank to just over 2 percentage points.
- Twist: the Santa rally was negative three years in a row at the end of 2023, 2024 and 2025, and its gap to the baseline in the probability of rising shrank from 25 percentage points in 1950–1989 to 9 in 2006–2025.
- The small-cap January effect has virtually disappeared. The Russell 2000's January excess return (vs. the S&P 500) is +0.09 percentage points in 1988–2005 and +0.02 in 2006–2026, and the medians are both negative.
- Year-end patterns are a filter, not a trading signal. Use them only to adjust the timing of a plan already made, and give them weight only when they overlap with other evidence such as trend, earnings and taxes.
A pattern that comes once a year, like the turn of the year, had only 76 samples. In Lesson 4, we look at why the turn-of-the-month effect, which comes around once a month, disappeared, and why the presidential election cycle, which comes around once every four years, still remains.
Quiz5-question quiz
Pick an answer and it's graded right away with an explanation.
Q1. Which period's return does the Santa rally describe?
These are the 7 trading days defined by Yale Hirsch in the Stock Trader's Almanac. In 1950–2025 the S&P 500 rose in 76% of 76 cases.
Q2. The Santa rally's probability of rising was 82.5% (baseline 57%) in 1950–1989 and 70% (baseline 61%) in 2006–2025. What is the most fitting interpretation?
A pattern's strength is read as its gap to the baseline. The interpretation that fits the data is that it remains but has become smaller.
Q3. How were the other 11 months in the 30 years from 1951 to 2025 when January fell?
The expected return was just lower than in years when January rose (+12.07%); the average was still positive. So don't use 'January down = sell signal.'
Q4. Looking at the Russell 2000's January excess return (vs. the S&P 500), what state is the small-cap January effect in?
It was +0.09 percentage points in 1988–2005 and +0.02 since 2006, so at the index level it effectively no longer remains.
Q5. Which way of using year-end patterns in investing best fits the principle of this lesson?
Seasonality is a tilt in probability, not a prediction, and it is a filter, not a standalone signal.
NextTerms · Next lesson · Related tools
The tendency of the US stock market to rise over seven trading days: the last five trading days of December and the first two of January. The S&P 500 rose in 76% of the 76 instances from 1950 to 2025, but the share of up periods for any seven trading days (the baseline) was also 58%.
January barometerJanuary barometerA rule of thumb that in years when the index rose in January, the average return for the remaining 11 months was also higher. For the S&P 500, the remaining 11 months averaged +12.07% in the 45 years from 1951 to 2025 when January was up and +2.07% in the 30 years when it was down, but for 2006–2025 it was +10.91% vs. +8.98%, so the gap shrank to a little over 2 percentage points.
January effectJanuary effectA pattern in which small caps returned more than large caps in January. It was first documented by Sidney Wachtel in 1942. The Russell 2000's January excess return over the S&P 500 was +0.09 percentage points in 1988–2005 and +0.02 percentage points since 2006, so at the index level it has virtually disappeared.
Excess returnExcess returnAn asset's return minus the return of a benchmark. For example, if the Russell 2000 returned +3% in January and the S&P 500 returned +2%, the January excess return is +1 percentage point.
BaselineBaselineThe 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.
Probability of a rise (win rate)Win rate, % upThe 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.
Tax-loss sellingTax-loss sellingSelling stocks that have lost money near year-end to lock in the loss and reduce that year's taxable gains. It is often cited as a cause of year-end weakness and the rebound at the start of the year.
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 testThis 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.
Put what you learned to use right away
Overseas Stock Capital Gains Tax Saving Simulator→Lesson 1 · The baseline and four numbers→For educational purposes only; this is not investment advice. Past statistics do not guarantee future returns. The S&P 500 was calculated directly from Yahoo Finance ^GSPC daily closes (January 1950 to October 2026) and the Russell 2000 from ^RUT daily closes (from September 1987). Price-index basis, excluding dividends, trading costs and taxes; retrieved 2026-10-09. Santa rally = last 5 trading days of December + first 2 trading days of the following January (vs. the prior trading day's close). The January barometer's 'other 11 months' = end-of-December close vs. end-of-January close.