Stats › Intraday and market microstructure › Overnight returns
Overnight versus intraday returns
Split a trading day in two. The close to open leg, when the exchange is shut, and the open to close leg, when it isn't. Add each leg up separately over twenty or thirty years. You get one line that goes up and one line that mostly doesn't.
That is the overnight return puzzle, and it's one of the few market anomalies that has survived being written about repeatedly. I put this page together because the claim gets repeated constantly with no numbers attached, and when you go and find the numbers they're both stronger and more awkward than the summary version suggests.
So here is what the actual papers measured, over which windows, and what happened when people tried to trade it. Seven sources at the bottom, all dated. The counter-argument gets its own section because it deserves one.
Across S&P 500 stocks from 2000 to 2022, the overnight return minus the intraday return averaged 2.75 basis points a day, about 7.2% a year, with intraday returns flat to negative. In QQQQ from 1999 to 2006 the close to open risk premium was +23.7% a year against -23.3% open to close. In E-mini S&P 500 futures from 1998 to 2019 the overnight session earned 2.6% a year out of a 4.3% close to close total. And yet a strategy that goes long the best overnight hour had a Sharpe of 1.1 before costs and -0.5 after bid-ask spreads, and the two ETFs built to harvest the effect were liquidated 13 months after launch.
Members only
The rest of this page is for members
Below this point there are 8 sections, 1 chart, 1 table and 7 named sources, roughly 2050 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.