AAPL overnight vs intraday return decomposition (last ~3 years, RTH 09:30–16:00 ET)
The headline is blunt: over the roughly three-year window (June 30, 2023–June 30, 2026) AAPL’s cumulative gain came during regular trading hours, not while markets were closed — the intraday leg compounded to about +68.28% while the overnight leg was −10.12%, leaving a close-to-close gain near +51.25%. That pattern runs counter to the “stock does its work at night” thesis.
This report decomposes AAPL minute bars into RTH open (09:30)→close and close→next open, compounds each leg across 751 trading days, and attributes total log-return to overnight vs intraday moves. The charts, full statistics and inference tests below show the calculation and the robustness checks that lead to the finding.
For AAPL over the past ~3 years, is the whole uptrend earned while the market is closed — does compounding only the close-to-open (overnight) moves capture the bulk of AAPL's cumulative gain while the open-to-close (regular-hours) return nets out flat-to-negative? Thesis: overnight holding accounts for essentially all of AAPL's advance while the intraday session is dead money, so the stock does its real work while you're asleep and fades during the hours everyone actually trades it.
How this was measured
Converted AAPL minute bars to New York time when available, filtered to regular trading hours (Mon–Fri, 09:30–16:00 ET). Defined daily RTH open as the 09:30 minute's OPEN and daily RTH close as the last minute's CLOSE before 16:00. Computed overnight return as open_t/close_{t-1}−1 and intraday return as close_t/open_t−1, alongside close-to-close total. Compounded each leg across the window and used log-return sums to attribute the share of total advance explained by overnight vs intraday.
The key numbers
Reading the numbers
Over 751 trading days, intraday (open→close) compounding produced about +68.28% while overnight (close→open) lost about −10.12%, leaving a close-to-close gain of +51.25%. In plain terms: the bulk of AAPL's cumulative gain in this window came from intraday moves, not overnight.
The charts
This cumulative-return line chart shows three separate paths: the intraday line ends at +0.6828, the overnight line ends at −0.1012, and the total (close‑to‑close) ends at +0.5125. Look at the growing gap between the intraday (top) and overnight (bottom) lines — intraday gains more than offset overnight losses. That pattern directly contradicts the 'all the work happens overnight' thesis: most of the upward drift is occurring during regular trading hours.
The end‑of‑window bar chart makes the same point in one snapshot: Overnight = −0.1012, Intraday = +0.6828, Total = +0.5125. Notice intraday alone is larger than the total because the negative overnight leg shaved off returns; the intraday leg is the dominant contributor to the net close‑to‑close gain.
The combined daily‑return histogram (n=1,502 observations) is tightly centered near zero (mean = 0.0004) with a range from −0.0942 to 0.1564, indicating many small moves and occasional large swings. The intraday leg is positive slightly more often (53.26% vs overnight 50.87%) and has a positive mean (headline mean intraday = 0.0007913 vs overnight mean ≈ −0.00008318). Statistical tests show no strong evidence that either mean is different from zero at conventional levels (overnight p=0.833, intraday p=0.1249), so while intraday compounding produced the large cumulative gain, average daily leg returns are small and not clearly significant by these t‑tests.
Calendar-year compounded returns within window
| year | n_days | overnight_cum | intraday_cum | total_cum |
|---|---|---|---|---|
| 2,023 | 126 | -0.0468 | 0.0449 | -0.004 |
| 2,024 | 252 | -0.0212 | 0.3352 | 0.3069 |
| 2,025 | 250 | -0.1165 | 0.2344 | 0.0905 |
| 2,026 | 123 | 0.0904 | -0.0229 | 0.0654 |
The takeaway
No — the data reject the "does its work overnight" thesis for this window. Compounded across 751 trading days, the intraday leg returned +68.28% while the overnight leg was −10.12%, giving a total close-to-close gain of +51.25%; on a log-return basis overnight actually shaved roughly 25.8% off the aggregate. Mean daily figures are tiny per day (overnight ≈ −0.008%, intraday ≈ +0.079%), and the fraction of positive days is similar (about 50.87% overnight vs 53.26% intraday). Formal tests are not ironclad: the overnight mean is indistinguishable from zero (p≈0.833 — roughly an 83-in-100 chance the observed mean is just noise) and the intraday mean is a suggestive lean but not conventionally significant (p≈0.125 — about a 12-in-100 chance). Practical takeaway: over June 30, 2023–June 30, 2026 AAPL’s cumulative advance came during regular trading hours, not while markets were closed; treat daily-mean inferences as suggestive rather than definitive.
The fine print
- RTH open uses the 09:30 minute's OPEN and RTH close the last minute before 16:00; auction prints can differ by a few basis points on volatile days.
- Half-days are included and shorten the intraday window; they can bias intraday averages unless filtered.
- One-sample t-tests assume IID returns and ignore heteroskedasticity, so p-values are descriptive rather than strict proof.