pairs trading clarification

Discussion in 'Strategy Building' started by finalsolution, Sep 28, 2003.

  1. Hi all,

    I was recently reading through "Professional Stock Trading" (which I have found to be an excellent book), and was quite taken by the pairs trading chapter. I generally prefer to swing trade, on a 2 to 20 day basis, so I was trying to rework the system from intraday to multi-day.

    My first question is, is this strategy even feasible on a longer scale than intraday? I can see some arguments that might suggest it couldn't be, but I haven't been able to convince myself.

    If indeed it is feasible, then I am having a somewhat difficult time determining how to make my spread calculation make sense. On an intraday basis, you simply reset the spread to 0 at the close of each trade, and trade appropriately the next day. But for a longer time frame, your spread period would seem to need to be longer, as suggested in the book. (5 days)

    But when using a 5 day period (for instance), and oddity arises. Let's say we start a system on day 1. On day 2, we create the spread comparing the current price to the close of day 1. On day 3, we compare the last price again to day 1. On so on. The comparison days would be

    Day 1 compare to day 1
    2 -> 1
    3 -> 1
    4 -> 1
    5 -> 1
    6 -> ???

    But on day 6, to what day should the comparison be made? Since we have a five day period, it is time to shift. But moving the comparison to day 5 (6 -> 5) ruins the continuity of the spread calculation. Simliarly, calculating like this makes little sense either:

    6->1
    7->2
    8->3
    9->4 , etc.

    I am completely new to this idea, so any help would be appreciated. I'm sure I'm misunderstanding something, but I've searched around quite a bit, and haven't had any luck.

    Thanks,
    fs
     
  2. Bob111

    Bob111

    PairsTrading.com

    http://www.pairtrading.com/index.php

    as far as i know-first one trade stocks only on daily basics. there is no intraday trades.
    the book that you read-is it help you to make any money?

    you don't have to think about odds and count days, all you have to do -is find a number of good pairs, which can be not correlated between each other,but making crossovers all time on timeframe that you will trade. that about it. :D:D:D

    good luck!
     
  3. When you say, find a pair that is crossing over, do you mean in terms of price, or spread? I'm assuming you mean spread.

    Perhaps what you are saying is that I should find two stocks, pick a starting date, and then never worry about resetting the spread. In other words, I will always compare to the same day in the past. So in terms of my first question, this would mean:

    2->1
    3->1
    4->1
    5->1
    6->1
    7->1 etc.

    In terms of the book, I can't say it has helped me earn any money because I just bought it one week ago. Also, I'm currently only paper trading a trend following system. So no, no money yet :)
     
  4. Bob111

    Bob111

  5. seisan

    seisan

    fs,

    Thanks to Bob's input, you've already extracted
    a good insight into evaluating pairs candidates.

    Since you just started this thread last night,
    are you aware of the successfully ongoing
    pairs trading thread also in this forum?


    Pairs Trading Strategy Model
    http://elitetrader.com/vb/showthread.php?s=&threadid=8533


    Good luck! :)
     
  6. seisan:

    Yes, I looked into the main thread, but didn't find the answer I was looking for, although there certainly was plenty of other useful information.

    I didn't quite grok bob's answer entirely :) I'm still not sure how to calculate the spread properly on a multi-day basis. I'm looking to program some chart indicators to visualize the current spread.

    Is the best approach to just have a stationary comparison point in the spread calculation?

    Thanks again,
    fs
     
  7. Bob111

    Bob111

    i think i did explain everything on tread mentioned above. spread can be calculated in number of ways. i did calculate % change and then-plote difference in % change. not money or price as others do.
    i believe-there is some pictures of it too.
     
  8. seisan

    seisan

    fs,

    Bob did assemble a complete answer for you...
    and is being a patient tutor-contributor...

    While you are re-reviewing his answer, keep the
    question corralled for a bit....

    I've called in the Royal Canadian Mounted Pairs Police :D
    to give you another angle to aid in your understanding!

    OK?
     
  9. Bob111

    Bob111

  10. #10     Sep 29, 2003