The Overnight Effect: Why Markets Move Most When They're Closed

You go to sleep. The exchange is shut. Nobody is buying, nobody is selling, the building is locked.

And yet, more mornings than not, you wake up to find the index has already moved. Not because of anything that happened on the exchange. Because of everything that happened while it was closed.

I came across a claim recently that sounded like the kind of thing you see on finance Twitter right before someone tries to sell you a course. A strategy that buys the Nifty at the close and sells it at the next morning's open, done every single day since the mid-1990s, would have turned a small amount of money into a small fortune. A cumulative return in the tens of thousands of percent.

My first reaction was the same as yours probably is right now: that sounds made up.

So I checked. And here's the part that actually surprised me. The real, sourced number isn't smaller than the claim. It's bigger. Not tens of thousands of percent. Over six hundred thousand.

The number, and where it actually comes from

A 2020 paper by researcher Bruce Knuteson, published on arXiv and titled "Strikingly Suspicious Overnight and Intraday Returns," studied exactly this split across twenty-one major stock market indices, the Nifty 50 among them, over roughly three decades.

The method is simple to describe even if the result isn't. Every trading day has two legs. There's the overnight leg, from the previous close to today's open, when the market is shut and nothing is technically "trading." And there's the intraday leg, from today's open to today's close, when actual buying and selling happens.

Knuteson's paper tracked what would happen if you compounded only the overnight leg, day after day, for three decades, and separately what would happen if you compounded only the intraday leg. For the Nifty 50, the overnight-only path came to +628,056%. The intraday-only path, over the exact same three decades, came to −99.86%.

Read that again. If you had only held the Nifty during actual trading hours, buying at the open and selling at the close, every single day, for thirty years, you would have been left with almost nothing. Nearly all of the index's real, actual, enormous growth happened while the market was closed.

Nifty, ~30 years: overnight-only = +628,056% | intraday-only = −99.86%

One detail worth sitting with: the Nifty 50's base value, on its actual launch date of 3 November 1995, was set at exactly 1,000. That's not related to Knuteson's study, it's just how the index itself was built. But it means the "starting point" in a lot of these overnight-effect claims lines up suspiciously well with a real, specific, checkable fact about the index, which is probably why the framing sticks.

This isn't new, and it isn't just India

The person usually credited with finding this is Ken French, yes, the same French from the Fama-French model that half of modern portfolio theory is built on. In 1980, as a young researcher working under Eugene Fama at Chicago, French asked a simple question: are stock returns spread evenly across the days of the week, the way the Efficient Market Hypothesis says they should be?

They weren't. French found that almost all of the market's gains happened Monday morning through Friday close, while the Friday-close-to-Monday-open stretch, the weekend, was on average negative. He called it the weekend effect, and it was the first real crack in the idea that markets behave the same way at all times.

In 1986, French and Richard Roll took it further, splitting every single trading day, not just weekends, into its overnight and intraday components. The pattern held there too. It's been reproduced enough times since that a former LTCM partner, Victor Haghani, called it "the grandmother of all market anomalies" in a 2022 paper. Nobody fully agrees on why it happens. Nobody seriously disputes that it does.

Why India specifically looks like this

Here's where it stops being mysterious and starts being mechanical, at least partly.

Before October 2010, the NSE opened cold. Trading started at 9:55 AM with no separate window to figure out a fair opening price first. Whatever information had piled up overnight, an earnings surprise, a US market move, a rate decision, got shoved into the first few minutes of live trading. On paper, a lot of what we'd now call an "overnight gap" actually showed up as an early intraday move, because there was no mechanism to price it in before the bell.

In October 2010, that changed. NSE introduced a pre-open session: a 15-minute window from 9:00 to 9:15 AM, split into an order entry phase, a price-discovery phase where the exchange runs a call auction to find one equilibrium opening price, and a short buffer before regular trading begins at 9:15. Instead of opening cold, the market now negotiates its opening price before anyone can react to it live.

Zerodha's own research team ran the Nifty data in two windows, 2000 to 2010 and 2011 to 2025, and the shift is not subtle. In the first window, overnight returns were mostly flat and intraday did most of the work. In the second, it flips hard: consistent gains overnight, consistent losses intraday. The same underlying information is arriving either way. What changed is which side of the 9:15 bell it gets priced into.

The honest part of this story is that it doesn't fully explain everything. The Sensex shows the same overnight-heavy pattern going all the way back to 2000, even though the pre-open auction applies to it too and should have caused a similar shift. Zerodha's own writeup admits they don't know why. Neither do I. Some of this is genuinely still unexplained, and I'd rather say that plainly than pretend the pre-open story wraps it up completely.

Why the market doesn't just erase this

If this is real and well documented, the obvious question is why arbitrage hasn't traded it away. A few honest possibilities, none of them fully settled:

Overnight is riskier to hold through. Earnings, central bank decisions, and geopolitical news disproportionately land after hours. If you're exposed overnight, you can't react until the market reopens. Some researchers argue investors demand a small extra return for carrying that risk they can't manage in real time.
Global market linkages matter. A lot of the Nifty's overnight gap comes from cues set by US markets, which are still trading, or have just closed, while India sleeps. The "overnight" effect in India is partly just the daytime effect somewhere else, arriving with a time difference attached.
Institutions and retail don't trade the same way at the same times. Research on order flow shows individuals trade more near the open, while institutions with genuinely large orders spread their activity through the day and concentrate more near the close, when liquidity is deepest. That timing mismatch alone can bend how returns land across the day.

So should you just buy the close and sell the open?

This is the part where I'd be doing you a disservice if I let the number speak for itself and stopped there.

First, "close" is not one clean number in India. There's the last traded price at 3:30 PM, and there's the official closing price, which NSE and BSE actually compute as the volume-weighted average of the last thirty minutes of trading. A large order at 3:25 can pull the official close away from what you saw flash on your screen at 3:30. Any backtest that assumes you can transact exactly at "the close" is quietly assuming a fill you might not get.

Second, the index itself isn't something you can trade. Nifty 50 the number is just a calculated value. Nifty futures are the actual tradable product, and futures carry real liquidity, spread, and impact costs that a clean index backtest never sees. A strategy that looks great on spot data can look very different once slippage is real money instead of a rounding error.

Third, and this is the one most people skip: holding overnight means you cannot exit if something goes wrong. During the day, if a stock or the market turns against you, you can sell. Overnight, you're locked in until the market reopens, however bad the news that broke while you were asleep. The same mechanism that produces the positive overnight drift on most days is exactly what exposes you on the days it doesn't. A strategy that wins on 90% of nights and gets flattened on the other 10% can still be a bad idea to run with real money and no plan for the 10%.

The Compound Move Take

The overnight effect is real. It's been documented since 1980, reproduced across decades and dozens of markets, and the Indian version has an actual mechanical story behind at least part of it. None of that is in dispute.

What's also true is that "real and documented" is a different claim from "easy money if you just show up." The gap between a clean statistic on a research paper and a strategy that survives contact with real transaction costs, real slippage, and real overnight risk is where most people who try to trade an anomaly actually lose.

The bigger lesson for me wasn't really about overnight returns at all. It was that the wild-sounding number I set out to fact-check turned out to be an understatement, not an exaggeration. That almost never happens. Most viral stats shrink when you check them. This one grew. Which is exactly why I checked it before writing a single word about it, and exactly why you should do the same with the next one you see.