Yes, Virginia, the Markets are Mean-Reverting

santa virginiaHere’s a stupid system. In fact, I call it the “Stupid 10 Days” system.

You look at the last ten days of trading. If the market at the closing bell is up from the market close of 10 days ago, you buy at the next open. You hold for 10 days, then sell at the next open. A classic momentum play. Wash, rinse, repeat. And remember, you have that uncle who owns a brokerage, so you don’t have to pay commissions. Start with $1000, invest as much as you can each time, and compound that puppy. Here’s what that system would get you with SPY since its inception:

momentummomentum drawdownThat’s just awful. Just when the system gets back on its feet, it gets smacked down again. “What a stupid system!” you’re thinking.

Let’s flip that stupid system on its head. If the previous 10 days were down, you buy the next open. You hold it for 10 days, sell at the following open, and repeat. $1000 start, invest as much as you can each time, and compound it all. Here’s what you get:

mean reversionmean reversion drawdownThat’s a WHOLE lot better. Not perfect mind you, as it still had some serious drawdowns. But this second system – mean reversion – is 15% less stupid than momentum. That’s science!

Compare to buy-and-hold:

buy and holdbuy and hold drawdownMean reversion, at least in a commission-free world, is better than buy and hold. And crazy-better than momentum. At least for SPY.

18 thoughts on “Yes, Virginia, the Markets are Mean-Reverting”

  1. Interesting study and back test. Q. How would results vary if diffrerent lookback/holding priod were used? For instance, if the past 5 days were down, then go Long and hold for 5 days. Or use 2days, etc for lookback and for holding period.

    Is there an optimum number of days for lookback, and holding, periodos?

    Thanks for sharing your work.

    1. If I split the test periods into two sections (1993-2009 and 2010-present), and run optimizations on both period separately, the ranges of 7-10 trading days and 37-39 trading days show the best results. it varies over time, but those two ranges both come up strong in optimizations of the two periods.

      What’s interesting is that this also improved the ‘system’ from my previous post, the M-W-F overnight trade. Not from a total-profit standpoint, but from a CAR/MaxDD standpoint.

      Also up for investigation would be looking at independent periods for entry and exit.

      Thanks for your comment!

  2. Thanks for your helpful reply. It explains why you used 10 days for the lookback and holding period.

    One other thought. Would you two large drawdown periods be limited if you used a stop trading kind of filter? For example if VIX was over 20 level, then stop trading.

    Interesting work. Thanks again for sharing.

    1. After I wrote the post, I did try implementing stops. It degraded performance, at least with my tests. But there may be a valid way of incorporating a stop that I haven’t tried.

    2. Jim, on further thinking, I realized you were suggesting a more sophisticated stop than the one I tried. I simply had an exit based on the percentage below the buy price…the usual stop in other words. You’re suggesting using either an exit or a filter to avoid entry based on VIX. That’s definitely worth looking at. I might also look at breadth indicators too.

  3. Some may look at these results and think mean reversion is better than momentum. But they are both effective over their appropriate time frames. Research in the 1980s and 1990s showed mean reversion works over the past month (and also 3-5 years), while momentum works over the past 3-12 months.

    1. Fred you’re right, and you bring up a good point. Markets are mean-reverting at some time frames, trending at other time frames (and this applies to intraday as well). I am guilty of being a little facile in this blog post about MR versus momentum. Probably because I tend to think mostly about shorter trades. 🙂

    1. Thanks Cesar, I remember that article and enjoyed it. It’s a ‘smarter’ system than this one. This system is more of a sketch than a true tradable system.

  4. You are comparing short-tern mean reversion to short-term momentum. Only an idiot would do that. You are not ashamed of yourself obviously and you think you have a story to tell when in fact you got nothing but paranoia.

  5. At the end of any day, the market can only do one of two things the next trading day. Either Mean Reversion, or Follow Through (momentum).

    Understanding which of these is dominant at any given time, is the most meaningful insight a swing trader could have for trading success.

    Only an idiot could ignore that reality.

  6. Well, I believe that if you extended the period to 20 days instead the results would be reversed. Momentum only works if you have enough time for the volatility in the market to even out a bit over time.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.