Hold Overnight?

I recently read the book “Inside the House of Money: Top Hedge Fund Traders on Profiting in the Global Markets“, which contained interviews with different global macro hedge fund managers in a similar style to the Market Wizards books. One theme that stuck with me after reading the book was the multiple references to “risk premium” which was used to describe the premium/return from holding an instrument for a specified time (ie overnight). Based on what I read in the book, I was curious if such a premium did exist and if so, how much of a premium it is.

My first stop in my research was this excellent post on TraderFeed where Dr. Steenbarger shows the difference in performance of the SPY in the overnight vs day session over the last three years. My next step was to see if a similar premium existed for individual equities as well. To test this, I created two simple scripts in Wealth-Lab. The first script would buy $10,000 worth of each Nasdaq 100 stock at the open and sell at the close for the previous year. The second script would also buy $10,000 worth of each Nasdaq 100 stock but this time buy at the close and sell at the next open, to isolate the “overnight premium”. Both were run in Wealth-Lab without any slippage or commission taken into account.

Here are the results for the “daytrade” script that buys the open, sells the close:

Here are the results for the “overnight” version that buys the close and sells the next day’s open:

What becomes obivous from these results is that over the past 12 months, a period in which the Nasdaq 100 performed very well, over 2/3rds of the total profit in the individual equities came from holding overnight. Furthermore, the profits are much more consistent and the maximum drawdown is a fraction of the daytrading script’s drawdown. Now obviously this is a relatively small sample size, but when combined with the above findings from TraderFeed, I start to think there could be something here.

What i found most suprising was when I looked to see how valid the often repeated axiom of “holding overnight is too risky” fared in this test. It turns out that the average volatility of holding overnight is actually less then holding it from the open to the close. This can be seen in the Average Profit % and Average Loss % which is 1.35% & 1.30 %for the daytrading script and only 0.67 %and 0.62% respectively for the overnight script. Therefore, the average Nasdaq 100 stock on up days moves 1.35% intraday and only 0.67% overnight, while having moves of -1.30% and 0.62% on down days.

I also investigated the relative # of large (> 10%) winners and losers between the scripts. On the overnight scripts, this are mostly gaps up and down related to earnings or other news related factors. Here are the results:

* Day trading script - 5 losers > 10%, 9 winners > 10%

* Overnight trading script - 22 loser > 10%, 24 winners > 10%

Obviously there will be more large winners and losers when holding overnight, but the ratio is not too far off for this small of sample size.

To help provide a visual example of the difference between the two scripts, I created this chart of GOOG for the last year. It graphs the previous 12 months and is normalized to start at a value of 0. The green line is the # of points gained by GOOG overall. The blue line is the # of points gained by the overnight script (buy each close, sell next open) and the red line is the # of points gained by the daytrading script (buy each open, sell each close).

This is still very much a research project work in progress, but I felt the results were already interesting enough to share with all of you. I look forward to any comments or emails with ideas of if this premium is really persistent and if so, how it could be used.

- John