On a recent morning the U.S. market looked calm from the outside — a low volatility reading, a steep, healthy yield curve, an index barely changed. But that calm was a veneer stretched over a night that had been anything but. While New York slept, Asia sold its chipmakers hard: reports had the Kospi down roughly 8%, SK Hynix off around 10% after record profit still missed forecasts, and Nvidia sliding about 5% in the pre-market on renewed financing worries. A U.S.-only view would have started the day blind to the single most important thing that had happened.
That gap is exactly what the overnight tape is for. By the time the opening bell rings in New York, the Asian session has already closed and Europe is well into its day. Those markets carry information — a chip-demand scare, a currency move, an earnings miss halfway around the world — that frequently is the reason the U.S. tape moves at 9:30. Ignoring it doesn't make it go away; it just means finding out late.
How much actually happens while the market is closed? A decade, measured
"Overnight matters" sounds like a slogan, so let's measure it. We took ten years of SPY daily data and split every 24 hours into two sessions: the overnight session (yesterday's close to today's open — the 17½ hours when you cannot trade the regular tape) and the day session (open to close — the 6½ hours when you can). Then we asked a simple question: where did the returns actually happen?
Over this particular decade, the closed market out-earned the open one: the overnight legs compounded roughly twice what the day legs did, even though nobody can trade a closed exchange. This "overnight effect" is a well-documented curiosity in market research, and the honest caveats matter: it is a description of one decade, not a strategy — capturing it would mean paying a spread twice a day, and the split flips around in some years and regimes. But the reading that survives every caveat is this: a large share of what the market does, it does while it's shut — landing in one untradeable jump at the open.
Most nights, that jump is small. Some nights, it's the whole week:
| Morning | Overnight gap | What happened while the market was closed |
|---|---|---|
| Mar 16, 2020 | −10.4% | COVID-19 lockdowns spread across the U.S. and the Fed slashed rates to zero in an emergency Sunday meeting — the market still opened limit-down |
| Mar 13, 2020 | +6.0% | Overnight rebound after the worst single session since 1987, as fiscal and emergency responses took shape |
| Aug 5, 2024 | −4.0% | The yen carry unwind: Tokyo cratered overnight and dragged the world’s opens down with it |
| Nov 9, 2020 | +3.9% | Pfizer’s first COVID-19 vaccine efficacy results, announced before the open |
| Nov 10, 2022 | +3.7% | A cooler CPI print hit at 8:30 a.m. — an hour before the bell |
| Apr 3, 2025 | −3.4% | Sweeping new U.S. tariffs, announced after the previous day’s close |
Look at the right-hand column: pandemic lockdowns, an emergency rate cut announced on a Sunday, a tariff regime unveiled after the close, a CPI print released an hour before the bell, and a carry trade that unwound in Tokyo — the one we walked through in our yen-carry post-mortem. None of these waited for New York's opening bell, and no U.S.-hours trader got to react at yesterday's price. The overnight session isn't dead time; it's where the decade's biggest single moves lived.
So how does an engine read a market that's already shut? From the news wire — the actual session that just closed — not from a proxy. It's tempting to reach for a U.S.-listed Asia fund as a shortcut, but those funds only trade during U.S. hours. At dawn in New York they still show yesterday's close, so they hand you a day-old signal dressed up as a fresh one. The live session lives in the headlines, not in a stale ticker. Read the wire, and you're reading today; read the proxy, and you're reading the past.
On that particular morning, the read mattered. A domestic backdrop that looked mildly constructive on its own numbers was tilted back toward neutral to account for the offshore shock — a concentrated scare about the durability of the AI-capex trade, concentrated in the chip complex. And the tape confirmed it: the U.S. market didn't hold its calm; it followed the overnight signal down. The lesson isn't that overnight moves always spill over — sometimes a foreign wobble stays foreign. It's that a quiet domestic open is not, by itself, the whole story, and that reading the world in the order it actually trades — Asia, then Europe, then New York — is often the difference between being early and being surprised.
What our seal does with the gap
This is also why our engine's schedule is built the way it is. Every trading day, the four strategies' decisions are sealed before the opening bell — after Asia has closed and the overnight tape has been read from the wire, but before the U.S. gap can be traded. Two consequences follow, and both are deliberate. First, a brand-new position is priced at that day's open: whatever the market gapped overnight, the engine doesn't get to claim it — it enters after the jump, at the same price any open-hours participant could have had. Second, every position it already holds carries the overnight session in full: our published returns include every gap, friendly or hostile, from entry open to exit close. The overnight tape informs the decision before the bell; it never inflates the score after it.
1. Over the past decade, a large share of SPY's returns arrived between the close and the open — in a jump nobody trading regular hours could catch. The biggest single moves of the era were overnight gaps, not intraday slides.
2. Read the session that just closed from the news wire, not from a U.S.-listed proxy that still shows yesterday — a stale ticker is a day-old signal dressed as a fresh one.
3. A clean record prices decisions at the open and carries every overnight gap it holds through. The overnight split is a documented historical pattern, not a strategy and not a prediction.
Informational and entertainment content only. Not investment advice. Session-split and gap figures computed from Tiingo dividend-adjusted SPY data; historical patterns guarantee nothing about the future.