Overnight vs Daytime Markets: Two Very Different Animals


I’m going to do a series of posts (inspired by Woodshedder) on stock market performance in the daytime (open-to-close) vs overnight (close-to-open) sessions.

As we’ve shown previously, the overnight and daytime markets are very different animals, and understanding that difference is of uber-importance to active traders.

Since at least 1993, bull markets have tended to play out overnight and bear markets in the daytime.

[logarithmically-scaled, growth of $1]

To illustrate, the graph above shows two fictional traders.

The first (in green) goes long the SPY (S&P 500 ETF) at every close and sells at every open. In other words, our first trader is only long the market overnight.

The second (in red) goes long at every open and sells at every close. Our second trader is only long in the daytime.

This is a proof of concept so results do not account for slippage or transaction costs.

Clearly, bull markets of the last 18 years have mostly played out in the overnight session (green). The daytime session (red) has been about 65% more volatile, but generally speaking, bearish.

This observation continued to hold true in our most recent bear market. The following graph shows the same test starting in October, 2007.

[logarithmically-scaled, growth of $1]

The overnight session (green) has recovered to where it stood before the crises began, while the daytime session (red) is still nursing a 40%+ drawdown.

This observation has enormous implications to active traders when deciding when to enter/exit positions, and flies in the face of conventional wisdom about the dangers of holding positions overnight.

But that really isn’t what this series is about…

I want to expand further on Woodshedder’s post looking at how short-term mean-reversion (or the tendency for very recent price movement to reverse course) relates to the overnight and daytime sessions.

Apologies for the anti-climatic post. This was meant to be a primer for those who’ve never heard of the peculiar discrepancy between the overnight and daytime.

As always, more to follow.

Happy Trading,

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.

23 Responses to “Overnight vs Daytime Markets: Two Very Different Animals”

  1. 1 Frank


    thanks for sharing.

    Although I absolutely agree with respect to the overall conclusion (overnight – positive, daytime – negative), you will probably come to a different conclusion digging deeper into the different time frames of the day session.

    The first hour of the sessions is almost as profitable (with respect to the probability for trading higher during the first hour, and the respective profitability like the median trade as well) as trading overnight (and closing the trade right at the open). The middle part of the session (start of last hour vs. end of first hour) shows a slightly positive tendency as well.

    Historically you would’ve achieved superior returns if you would’ve been invested from a session’s close until the start of the last hour of the then following session, no position at all during the last hour of a session.

    It is the last hour of the sessions which historically shows a significantly bearish tendency, on average not only erasing all of then recent intraday gains, but turning probabilites and odds for the day session into the red.

    Stats (winning percentage | compound return) woud’ve been (SPY, since 1/1/1993):
    open vs. previous close: 55.08% | +510.31%
    first hour: 53.94% | +421.27%
    start of last hour vs. end of first hour: 52.17% | +111.16%

    start of last hour vs. previous close: 57.57% | +6617.88%

    last hour: 44.40% | -94.64%

    Full stats can be found here: http://twitpic.com/2t9mtp/full

    (compound return doesn’t not account for the initial invest of 100%, therefore simply add +100% in order to match compound returns on Michael’s figures)

    I’m looking forward to your next postings in this series.


    • Frank can you clarify what your column headings mean? What is ‘last hr vs open’ and ‘last hr vs prev close?’ Once I know these, I should be able to figure out the others. Thanks.

      • 3 Frank


        ‘last hr vs open’ represents the SPY’S performance between the start of the last hour (3pm) and the session’s open (on the same day).

        ‘last hr vs prev close’ represents the SPY’S performance between the start of the last hour (3pm) and the previous session’s close.


      • Great…thanks Frank

  2. 5 Sebastian Fainbraun

    I did this same study a while back when testing short exits on a system that I trade. Amazingly this had no effect on the short side but gave me a bit of a pick-up if I got out the next day at the open or at the close on the same day the signal is generated. You should try mixing this strategy with an MA crossover, say being only long overnight when you are closing above a 200 day MA.

  3. 6 Sebastian Fainbraun

    sorry meant that I got a pickup when I held the longs overnight as opposed to getting out right on the close.

    • 7 MarketSci

      RE to Sebastian: thanks for the comment.

      The observation isn’t meant to be a strategy. I would never suggest anyone trade long every overnight session (or anything to that effect). The point is to show that the observation exists for folks who weren’t in the know.


      • 8 Sebastian Fainbraun


        I wasn’t suggesting that you were. But many people do employ this kinds of strategies – what they do is buy overnight in trending markets and simultaneously buy cheap puts just in case of catastrophe overnight.

        Personally, I hold longs overnight after an exit signal for the same reason you stated in the post, I find that in the long run, you get nice upside moves overnight. One thing that I havent looked at and @Frank probably has is the effect of morning announcements on the overnight market.

  4. 9 Sebastian Fainbraun


    What are you using to generate that report? Is it generated by Metastock?

  5. 10 Frank (@TradingTheOdds)


    data courtesy of MetaStock, and for data import, testing, surveys and statistics I use MATLAB from MathWorks. But all of those tables and stats are self-programmed (there are no pre-defined reports like that).


    • 11 Sebastian Fainbraun

      Thanks, also what confidence interval do you use for your T-Scores to consider a strategy’s success not attributable to chance?

  6. Looks like a lot of people have run studies on this topic. In mine, it appears that for the last year (10/01/09 – 09/29/10) the SPY has advanced $4.48 during daytime sessions and $4.33 overnight, giving a slight advantage to the daytime sessions. For my study, the SPY started out at 105.43 and closed yesterday at 114.51. Seems like the market’s always gaming anything already commonly known.

    • 14 MarketSci

      RE to Phil: I wouldn’t draw such a strong conclusion. There have been multiple extended periods where the daytime session outperformed the overnight (see first chart), but the broader “trend” has always returned. I don’t see any reason to think this observation is waning in significance (yet). michael

  7. I’d be interesting in seeing a similar study performed with the ES which trades almost 24 hours. Instead of looking at two time periods (overnight, regular trading hours), many more time periods can be studies based on when Europe opens or an hour before the US opens (because that’s when lots of news comes out). I have 24 hour ES data from 1998-2008 but I’m not a computer wiz…playing with the data in Excel is a pain.

  8. Looking forward to your work on this Michael!

  9. 17 heywally

    Some great posts here and it’s good to see Frank from tradingtheodds back online.

    I’ve been wanting to do something with this idea ever since I found out about it several years ago but with some sort of ‘filter’ applied. The being above a certain moving avg idea seems like a good one but the problem with that is that it could cause you to miss some big overnight bounces from heavily oversold conditions.

    I am now manually pouring through 15 minute overnight charts.

    If I come up with something, I’d probably run it with the NQ futures and a 1% stop (with a generous limit from there) which has the advantage of probably/possibly getting you out of a position during some kind of overnight disaster that ends up limit down and/or closing the markets.

  10. How do dividends factor into this analysis?

  1. 1 Using the Market’s Clock and Calendar StockTwits U
  2. 2 Thursday links: cheap options Abnormal Returns
  3. 3 Trading TZA into October | Brute Force Attack
  4. 4 When Is the Best Time of Day to Buy Stocks? Crossing Wall Street
  5. 5 A Strong Opening — And Then What Crossing Wall Street

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s