This all started over the weekend, when I started wondering about a trade I had going. The trade hadn’t hit my profit target on Friday and so carried over through the weekend. I started wondering about day-of-week ‘seasonality’ and thought I’d bust out the old charts and see what’s up.
Let’s take a look at SPY from 2000 through the current date. If we buy on a given weekday at the open and sell at the close, what’s the tendency? Is one day of the week better than the others? Here are the averages for each day of the week, open-to-close:
Friday is apparently not our friend.
What about overnight? If we buy Monday at close, and sell Tuesday at the open for example, is there any significance there? Does holding over the weekend hurt us or help us? Here are the averages for overnight trades.
Holding Thursday night is also an unpleasant prospect, at least when it comes to averages. But does that hold true generally, or are there some big moves skewing the averages? Let’s make some simple systems that trade on these days. No commission (because you have an uncle who owns a brokerage and lets you trade for free…hey we’re pretending.). $10,000 per position (because hey, we’re pretending!). No compounding, because we’re tough and don’t need compounding to make us feel good at night.
Fridays do suck. Look at the bull market of 2003-2007. Friday couldn’t even get out of bed. It’s hard to draw conclusions about the other days of the week, however Wednesday spent the most time ‘above water’. Don’t you find it interesting though that Wednesday and Friday returns almost look like mirror images of each other? Here are just those two days (open to close), for emphasis:
Also, please note that the dates are approximate. Due to holidays, there might be fewer trading days for one day of the week compared to another. The equity curves are independent, and the date below the graph is taken from the Monday equity line. You can see that the ends of the graph don’t exactly line up, so there is probably shifting of up to 10-20 days or so over time.
Let’s now turn our attention to overnight trading, which has the potential for higher returns (according to our averages bar chart).
Wednesday night is profitable even during the dot-com bust, but then lags a bit after 2008. Monday night makes money during the 2008 crash. And Friday night (i.e. over the weekend) putters along nicely throughout all this: not too hot, not too cold, but steadily profitable.
What about Tuesday and Thursday? Yikes. Those overnight trading periods suck. They appear to especially suck during bear markets. Those days suffer the greatest losses in the dot-com bust and the 2008 crash.
Let’s propose a simple system where we go long at the close of Monday, Wednesday and Friday, and sell at the next trading day open. Here’s how the equity curve would have looked:
How about doing the same thing at the close of Tuesdays and Thursdays instead? Dismal….
It’s not that you can’t make money on a Tuesday or Thursday night. It’s just that in bad markets, these overnight periods seem to get hit especially hard. “Sh– happens on Tuesday and Thursday nights.”
I find this bizarre to say the least.
Is this all just curve-fitting? Well I can’t say for sure, since I didn’t do an in-sample/out-of-sample test. However there’s consistency through different time periods, which makes me think that this is not due to randomness, nor to a single period of unusual activity that skews the results. This is especially true for the Tuesday/Thursday overnight data. The dot-com bust, the crash of 2008, even the correction in 2010 and the recent correction…all are much more severe on a T-Th overnight system than they are on a M-W-F overnight system.
Does this mean you should never hold positions overnight on a Tuesday or a Thursday? I don’t think that’s true in a general sense. However in a turbulent market, it may very well be the case. Perhaps there’s even a system that could be devised to short the market on those nights during downturns. But I’ll leave that to you to devise.