Methodology: How We Calculate Seasonality, Earnings Reactions and Cycles

In short: every figure on SeasonalityCharts.com comes from daily closing prices, compounded, using complete calendar years only, and is recalculated after each trading session. Seasonality is measured three ways (by calendar date, by week and weekday, and by trading day of the month), earnings dates come from companies’ SEC filings, and multi-year cycles are tested against what random price movements would produce. All figures describe the past; none is a forecast or investment advice.

The data

  • Daily open, high, low, close and volume for each symbol, back to 1970 where available, adjusted for stock splits but not for dividends (so returns are price returns).
  • Every update is checked for gaps, duplicated days, frozen prices, bad prints and unadjusted splits before anything is published. Symbols whose data cannot be verified are left off the site.
  • Each page shows the date of its latest session (“data through”), and every chart, table and statement on that page is rebuilt from the same data.

Year-to-date returns and complete years

A seasonal chart lines up each year’s year-to-date return: the change from the previous year’s final close to each session, compounded from daily closes so that a 10% gain followed by a 10% loss counts as the 1% loss it really is. On weekends and market holidays a year keeps its last close, so every year is present on every date. Only complete calendar years enter the averages: the current year and a partial first year are shown as lines but never averaged, so they cannot distort the pattern.

Three ways to measure seasonality

Method 1

By calendar date

Each year's year-to-date return on the same date (Jan 2 … Dec 31), averaged over complete years. Best for date-anchored effects such as the January effect or tax-loss selling.

Method 2

By week and weekday

Sessions lined up by week number and weekday, so "week 14, Monday" compares like with like across years. Best for weekday tendencies and week-by-week drift.

Method 3

By trading day of the month

Sessions numbered 1, 2, 3 … inside each month. Best for flow-driven effects such as the turn of the month and month-end rebalancing.

How we calculate seasonality, earnings reactions and cycles

A pattern that appears in all three views is more robust than one that appears in only one. Each symbol page also compares the symbol’s average year with the S&P 500 (SPY) over exactly the same years, and plots the three methods on one axis.

Best and worst months, win rates and the next four weeks

  • Monthly returns run from the last close of the previous month to the last close of the month. The best and worst months are those with the highest and lowest average over complete years.
  • “Up in 26 of 39 years” (the win rate) counts the years in which a period ended higher. Read it next to the average and the median: one exceptional year can lift an average while the win rate stays near 50%.
  • Next four weeks is the return over the 20 trading sessions after today’s date in every past complete year. “Historically strong” means it rose in at least 60% of years with a positive average; “historically weak” means it rose in 40% or fewer with a negative average.

Heatmaps

The two monthly heatmaps show every year and month. The cumulative heatmap adds up open-to-close moves within the month (the trading-hours return, leaving out overnight gaps); the absolute heatmap is the change from the month’s first open to its last close. Comparing them shows whether a month’s gains came during trading hours or overnight.

Earnings reactions

Earnings release dates and reported earnings per share come from companies’ filings with the U.S. Securities and Exchange Commission (Form 8-K, Item 2.02, which exists since August 2004, and the XBRL financial data in 10-Q and 10-K reports). The reaction day is the first session that could react: the same day for releases before the open, the next day for releases after the close. Fiscal quarters follow each company’s own fiscal calendar. Analyst estimates are not used. Funds, indexes and companies that do not file these U.S. forms have no earnings section.

Multi-year cycles

For symbols with at least 20 complete years, monthly closes are compared with their long-term trend (a 10-year local regression of the log price). We test every cycle length from 2 to 8 years and compare the strongest one with 120 simulated random price paths that have the same persistence: the share of simulations producing an equally strong cycle is the chance it’s luck. Major peaks and troughs use a reversal threshold scaled to each symbol’s volatility, and declines are checked against 13 S&P 500 bear markets and crashes since 1968. More in our guide to stock market cycles.

Limitations

  • Prices exclude dividends, so long-run returns of high-yield stocks are understated.
  • Seasonal averages rest on a limited number of years; a pattern seen in 20 or 30 years can still be chance.
  • Market structure changes: patterns from the 1980s may not hold today.
  • Nothing here is a forecast or a recommendation. See the Disclaimer.

Frequently asked questions

What is stock seasonality?

Stock seasonality is the tendency of a stock, ETF or index to perform differently at particular times of the year, month or week. SeasonalityCharts.com measures it from daily closing prices by calendar date, by week and weekday, and by trading day of the month.

How is the seasonal average calculated?

Each complete year’s year-to-date return is compounded from daily closes, carried over weekends and holidays, and averaged across all complete calendar years on each date. Partial years are shown but never averaged.

What does the win rate mean?

The win rate is the share of past years in which a period, such as a month or the next 20 trading sessions, ended higher. It is shown as “up in X of Y years” next to the average and median.

Where do the earnings dates come from?

From companies’ filings with the U.S. Securities and Exchange Commission (Form 8-K Item 2.02 and XBRL data in 10-Q and 10-K reports), since August 2004. Analyst estimates are not used.

How often is the data updated?

After every trading session. Each page shows its latest session date, and every figure on the page is rebuilt from the same data.

Is seasonality a reliable trading signal?

No. It is historical context. Patterns rest on a limited number of years, can be chance, and can change. Use them alongside other research and see the Disclaimer.