SPY OpEx-week vs ordinary-week returns and daily ranges (last ~3 years)
The popular OpEx-hour and triple-witching lore promises wilder moves around monthly options expiration. In a ~36‑month SPY sample (1,096 days, 36 OpEx weeks vs 121 other weeks) that claim doesn’t hold up: the mean Friday‑to‑Friday return was 0.4213% on OpEx weeks versus 0.3613% on other weeks, and the Welch t‑test gives p = 0.87 — no evidence of a real mean difference. Weekly realized volatility and daily intraday ranges are similarly indistinguishable (p = 0.555 and p = 0.769).
Below you’ll find the full dataset, charts, and tests. I resampled minute bars to daily OHLCV, flagged weeks whose Friday is the third Friday of the month, and compared weekly returns, within‑week vol, and daily high–low stats using Welch two‑sample t‑tests. The detailed results are next.
For SPY over the past ~3 years, do monthly options-expiration weeks — the week containing the third Friday — actually deliver stronger or wilder returns than ordinary weeks, the way the 'OpEx pin' and triple-witching lore claim? Thesis: average OpEx-week returns are statistically indistinguishable from every other week and only the daily high-low range ticks up modestly, so the expiration mystique is mostly noise rather than a tradeable edge.
How this was measured
Resampled SPY minute bars to daily OHLCV over the last ~36 months. Defined weekly periods ending on Friday (W-FRI). Flagged OpEx weeks as those whose week-ending Friday is the third Friday of the calendar month (standard monthly options expiration; if Friday is a market holiday the weekly anchor remains that Friday date). Compared: (i) Friday-to-Friday weekly returns, (ii) within-week realized volatility (std of daily returns per week), and (iii) daily intraday range (high−low)/close on OpEx-week days vs other days. Differences are assessed via Welch two-sample t-tests (unequal variance).
The key numbers
Reading the numbers
Across 1,096 days covering 36 OpEx weeks and 121 other weeks, mean Friday-to-Friday returns are ~0.42% for OpEx vs ~0.36% for other weeks, but a two-sided Welch p-value of 0.86995 shows that difference is not statistically meaningful and intraday ranges also show no clear gap.
The charts
The box plot contrasts weekly Friday-to-Friday returns: OpEx weeks (n=36) have mean 0.0042129 with observed weekly returns ranging from -0.0299 to 0.0442, while other weeks (n=121) have mean 0.0036125 with a wider range from -0.0877 to 0.0606. Visually the boxes and whiskers overlap a lot and other weeks actually show the largest single-week extremes on both sides. A Welch p-value of 0.86995 means there's no statistical evidence of a different average weekly return for OpEx weeks, so the plot supports the thesis that OpEx weeks are not reliably stronger.
This box plot shows daily high-minus-low divided by close: OpEx-week days (N=173) have mean intraday range 0.0548225 with min 0.0043 and max 0.2437; other-week days (N=579) have mean 0.0562430 with min 0.0035 and a much larger max 0.9017. The two distributions overlap heavily and the biggest single-day swings come from non-OpEx days, not the OpEx group. A Welch p-value of 0.7694 confirms no clear difference in typical intraday range, so OpEx days are not consistently wilder.
Weekly return and volatility summary by group
| group | n_weeks | mean_weekly_return | median_weekly_return | std_weekly_return | mean_weekly_vol(std of daily rets) |
|---|---|---|---|---|---|
| OpEx weeks | 36 | 0.0042 | 0.0062 | 0.0189 | 0.0075 |
| Other weeks | 121 | 0.0036 | 0.0042 | 0.0204 | 0.008 |
Daily intraday range summary by OpEx-week membership
| group | n_days | mean_range | median_range | std_range |
|---|---|---|---|---|
| OpEx-week days | 173 | 0.0548 | 0.043 | 0.0525 |
| Other-week days | 579 | 0.0562 | 0.0441 | 0.066 |
The takeaway
No — OpEx weeks do not show stronger or wilder SPY returns over the past ~3 years; they are statistically indistinguishable from ordinary weeks. The mean Friday-to-Friday return was +0.42% on OpEx weeks versus +0.36% on other weeks (36 vs 121 weeks) and the weekly-return t-test gave p = 0.87, offering no evidence of a real mean difference. Within-week realized volatility averaged 0.75% on OpEx weeks versus 0.80% on other weeks (p = 0.555), so week-level volatility is essentially the same. At the daily level the average intraday high–low was 5.48% on OpEx-week days (N=173) versus 5.62% on other days (N=579), with p = 0.769 — again no meaningful difference. Given the large p-values (0.87, 0.555, 0.769) and only 36 OpEx weeks, this is not a signal you can rely on; it reads like noise or a coin flip rather than a tradeable edge. Practically: don’t bank on an OpEx-week mystique — any edge is tiny and statistically unsupported in this sample.
The fine print
- Only 36 OpEx weeks in the sample — tests are low power for small effects.
- When the third Friday is a market holiday the classifier anchors to that Friday-week; exact expiration-day effects could be muted.
- Results use Friday-to-Friday weekly returns; alternate window definitions (e.g., Mon open–Fri close) may change measured values.
- This ~36-month window spans specific market regimes; effects could differ in other periods or longer histories.