Most volatile days of the week

Discussion in 'Automated Trading' started by travis, Sep 11, 2006.

  1. travis


    I am doing a study on the last 70 years of Dow Jones to find out two things:

    1) What are the most volatile days of the week? Monday, Tuesday...?

    2) What are the weekdays that are more likely to end up with a gain, and viceversa?

    I asked here, because my study is all statistical, and elsewhere they wouldn't appreciate it.

    For point 1) I used absolute gains/losses, and I found out that pretty much all days have the same volatility - it surprised me a lot, because I thought that mondays would be the most volatile. Do you have any different findings on this questions?

    For point 2), I added gains (or losses) for every weekday, during the 70 year period, and I found out that there is a big difference, but only in the past 20 years, and none if we consider the entire 70 years period. (Also for point 1) I analysed the last 20 years, and found no big difference between days' volatility). The average yearly difference in gains for the past 20 years is as follows:

    Monday: 6%
    Tuesday: 3%
    Wednesday: 2%
    Thursday: -1%
    Friday: 0%

    This could be exploited, if it is indeed exact, in a trading system, by making requirements for long trades easier early in the week.

    Beware, friday's +6% vs thursday's -1%, on the 20 year period means that, in total, we made 120% on fridays and -20% on thursdays. I don't think it's a difference that can be ignored, especially since it's happened constantly for the past 20 years.

    Another interesting thing I must mention is that for the years before 1988, up to 1930, just the opposite is true - monday's have the biggest losses, and wednesdays have the biggest gains. Assuming this is correct, do you have any idea why this could be? Electronic trading? New types of traders in the market after 1988?
  2. travis


    I was forgetting to add something important about the daily gains. It's necessary to compare monday's total gains to monday's total absolute move. If the total absolute move was 6%, then we could just go long on mondays and make money, each and every time, assuming this 20 year trend continued.

    Yet, monday's absolute total move for those 20 years, is 36%, just like on all other days of the week.

    Now, this means that each monday of the year, in total, we go up 21% and down -15%, leaving a gain of 6%.

    On the other hand, on each thursday of 1 year, we go up 19%, and down 17%, leaving a loss of -2%.

    Are these extremes far enough for us to be able to exploit these data in a system? I don't know yet. Certainly it's better than nothing, but is it enough to prevent our system to go short on mondays? Certainly it's not enough to prevent our system from going long on thursdays, being the difference so small.

    But 21% vs -15% for every monday of 20 years, assuming it will continue to that enough to tell your system to only go long on mondays, since they are more likely to be up days?
  3. The most volatile day of the week is Tuesday. Equally volatile are Thursday, Monday, Friday and Wednesday....
  4. travis


    Yes, thanks for sharing your opinion. I wish you shared some data backing it up, too. So far it sounds like an opinion. I am not saying it's not true.

    Here's my file, if anyone wants to look at it.
  5. You might want to take that file down. I was looking at it and it appears that you know the High, Low, and Close in the year 2028. If I had that information I dont think I would be sharing it. Kind of like in back to the future II ( i think) were Biff gets the sports almanac from the future. Its like its not even fair.:p

    --I'm just joking around, im sure its just something with excel, doesn't like dates for some reason.
  6. You are in serious need of the application of my famous 2400-year-old proven method to help you answer your own questions based on what you already know.
  7. travis


    Well, sorry, no one download that file. It has my predictions for the future in it!!!

    Just kidding. It's because I wrote the file on a computer set to European format dates, and it got all screwed up as usual.
  8. That's why I only trade on the weekends. Keeps my PnL pretty level.... :p :p :p

    I would have guessed Monday is the most volitile, simply because there are three days of news to accumulate for one trading day. Shows what I know! Also I would have assumed Friday was the least, because all the fat-account traders are up in the hamptons or something..... Maybe they take monday off too, evening it out.

    - TNG
  9. Pabst


    My best guess is that Monday's are reflective of the macro-trend. Makes sense from a psychological perspective. If the market's in a bull mode many investors and managers may be pining all weekend that they "must get in" as soon as trading begins for the new week. Likewise in a bear market the urge to puke is probably a worrisome emotion leading into Monday's action.......
  10. travis


    Pabst, the problem with what you are saying, which is also something I wondered about, is that I didn't see any correlation in my data between down years and mondays on those years - they should have been "down" mondays, but they weren't. The down mondays are evenly distributed in all situations and trends. For example, in the years 2001-2003 the mondays are not so down as you would expect, not any more than all other days, actually less.
    #10     Sep 11, 2006