Another N-days down and then what happens thread

Discussion in 'Strategy Development' started by SMI, May 17, 2012.

  1. SMI

    SMI ET Sponsor

    Given the sell off over the last week I wondered what happens after 5 consecutive days where the close is less than the open on the ES. So, I wrote a Tradestation strategy to test this. The results are interesting.

    Buy after 5 consecutive days down and hold for anywhere for 1 day to about 3 weeks and the aggregate result is a loss over the last 10 years. But, after that time, the results start to go positive in both dollars and % wins. The % win goes from below 40% to as high as 80% with the sweet spot being somewhere between holding for 60-90 trading days (3-4 months).

    A screen shot with the results here is here: http://screencast.com/t/hkCkMlPwca. This is a large image so you will have to click on it twice to open it and blow it up so that it's readable (at least this is the way it would work in internet explorer).

    What this tells me is that we're most likely close to, but not quite at the lows yet - patience, patience, patience. Over the next three weeks we are going to go lower. BUT, for a buy and hold investor (or for an IRA), the next three weeks could be a buying opportunity to sit on this stuff for 60-90 days.

    The full Tradestation code for you to play with and a pre-configured workspace is located here:
    https://www.structuredmarketseducation.com/Default.aspx?TabId=138
    Scroll down to the middle of the page to the Strategy section.
    (Note: Free registration required to get to the files.).

    Here are screen shots of the performance report and annual returns when buying and holding for 60 trading days.
    http://screencast.com/t/EuAL3sV8Pjej
    http://screencast.com/t/mh9gUlLZSu

    Note that the sample size is SMALL - only one or two trades a year and a couple of those years were negative.

    Hope some of you find this useful.
     
  2. SMI

    SMI ET Sponsor

    Ok, so we might have SIX consecutive closes down today. Therefore, I re-created yesterday's code to do 6 days down. Only about 9 instances over the last 10 years - about 1 per year until last year which had 3 of them.

    You can see the year by year results here: http://screencast.com/t/Pa6QNsKtIhBC

    The interesting thing here is that the best results are shorter term in nature compared to 5 days down - the sweet spot is holding for about 20 trading days or so. Of course, with 9 instances it's not exactly a statistically valid sample - more like a tendency you can trade around. Still, with 10K-17K on 8 trades and a 70-80% hit rate in the sweet spot, it's something to pay attention to. I'm not sure I'd want to be short going into the weekend if it looks like we might close below the open again...

    I posted the Tradestation code and workspace for download (registration and login required): https://www.structuredmarketseducation.com/Default.aspx?TabId=138
     
  3. Its strange to see these kinda of research. It seems you arent thinking hard enough.

    What if I used a 2.5% standard deviations or rven 3 for this kinda of events (where the last day pop or drop touched the 3 stdev mark - being 98% confident) and see if its ok to buy/sell after the 3rd day (probably after some market news digestion.

    This is more likely be a stronger case as we know markets can drop consectively for 6- weeks, with 3 days of drop every week (that also suggest a possible 5 day consec drop too depending on how the news are released 1 by 1)

    Try harder... :)
     
  4. SMI

    SMI ET Sponsor

    Thanks for your response. Your ideas are valid of course. But, for the purposes of this thread I wanted to keep things simple. By expressing the idea the way I did, anyone can pull up a basic chart and see the idea in action without having to put other indicators on the chart or know what a stddev is.

    I have noticed though that HOW a market gets to an extreme is as important as the fact that it got there, maybe even more important. An extreme that is reached via backing and filling can have a diferent probably outcome than one that is reached via a fast hard drive down (or up). So, 5 or six consecutive closes can have a different probably forecast than just a pure stddev move.

    I am sure many others will find your ideas useful. Thanks for your constructive critiscim and willingness to share them.