Pair Trading Article

Discussion in 'Strategy Building' started by Babak, Feb 13, 2003.

  1. The RD/SC article got me going with pairs trading and I played RD/SC quite a lot last year. It's never been the same since RD was removed from the S&P 500. The problem I had with the article (and I'm doing this from memory) is that the buy/sell points and backtesting were based on closing prices. If you decided to get in at the next day's open, you found that the spread had narrowed. I don't believe this was incorporated in the author's calculations.

    I also got the article from Harvard (not Yale) that the author referenced. It was quite helpful to understand the fundamentals of why the RD/SC ratio would change when the ratio of the profit split was essentially fixed.
    #11     Feb 14, 2003
  2. Babak


    you are correct, the testing was done as of close and the exit was also a close

    as for the Yale/Harvard article (whichever one it really is) do you have a link please? TIA
    #12     Feb 14, 2003
  3. There's no link. Harvard doesn't want to give anything away for free. I bought it from Harvard Business School Publishing. It is article 9-296-077 from July 29, 1997 and is titled, "Global Equity Markets: The Case of Royal Dutch and Shell". I don't recall the cost but probably only $10 or so.
    #13     Feb 14, 2003
  4. RD/SC stopped being a good pair because of all the attention given to it over the years...too many arbs mean the good spread prices are gone by the time you click the mouse.
    #14     Feb 14, 2003
  5. MrDinky


    I plugged in a couple stocks and got this error message (which I thought was pretty funny...)

    The system returned the following error in response to your request:

    Table 'bright_trading.BKX-X_eod' doesn't exist getting symbol data

    #15     Feb 14, 2003
  6. 1] Regarding the normalization equation, define delta!
    (I think delta= {RD/SC} - 30MA(RD/SC) but I'm not sure because he doesn't define it explicitly)

    I believe he's using {SC - (RD * 30MA(SC/RD))} to arrive at a premium or discount for SC. Positive delta represents a premium over "fair value" - SC is a over-priced.

    2] In the Pairs Relationship box (upper left hand corner) I think he puts RD instead of SC in the last sentence of a paragraph. It caught my eye because he mentions the ratio of RD to RD which makes no sense!

    He says they're traded on the AMEX as ADR's. They may have been at one time, but they're both big board stocks now. RD trades as common stock. SC trades as an ADR (based on STT listed in London). The 6:1 ratio must be 6 shs of STT represent 1 ADR share of SC (enough initials for you there?). I haven't confirmed the 6:1 ratio.

    3] The July 98 data that he provides as samples do not jive with my July 98 data (the prices of SC and RD are WAAaay off -- was there a split/reorg ?!?!)

    I checked via TC2000 and eSignal and Figure 3's closing prices look correct. I looked at unadjusted and didn't see a difference.

    4] PnL diagram doesn't make sense! Look at the right hand corner and notice there is a bunched up group of + bars. This should make the cumulative PnL line go up. But it goes down!

    Good eye! I thought perhaps dividend related, but SC paid out divs a few times through the study. Checking the spread I would think perhaps a sign error. Looks like he would have taken at least one loss during that time period.

    I agree with an earlier post in that this pair was traded into oblivion. It was tough intraday, and just slow on any other timeframe. About that same time a lot of the other pairs were getting tougher (other folks around me trading them - hearing their curses). Personally, I think the M&A arbs had to find something to do when that fountain dried up. That meant a lot of fish in the same pairs pond.

    However, I think pair trading has some potential. I've off and on for the last two years worked on coming up with some sort of convergent / divergent trigger, something to hint at which way it's likely to go. I've just got to convince myself there's an edge to it.

    Good luck with it
    #16     Feb 14, 2003
  7. I pair traded full time for almost a year and learned what works and what doesn't--the hard way.
    #17     Feb 14, 2003
  8. Good work and good patience, NYSE! I've finally gotten to the point of holding my own after 1-1/2 years of actively trading. I started trading in '97, but I don't really count the first few years, even though I made good oney. As they say, any idiot and E-Trade back then..
    #18     Feb 15, 2003
  9. Babak


    oh...ok! I think I understand what he is doing (atleast what you're saying he's doing).

    He's looking at this interms of comparing SC's actual close to what SC's theoretical close should have been. Then he's taking that difference and normalizing it.

    But what I thought he was doing was looking at the ratio and comparing it to what is should be. Then taking that difference and normalizing it.

    I'm not sure what the exact difference can signify but I kind of like my way of doing it better because what I'm doing is comparing the two of them together. Rather one against the other.

    I'm not sure if its right (?) but would appreciate your opinion on it. Any quant jocks want to give their take on this? (man?)

    ps in my previous post I said RD/SC when I meant to say SC/RD
    #19     Feb 16, 2003
  10. I'd rather go by the ratio of the two as well. However, I did like his normalizing approach.

    I did a bit of work with the SC/RD daily data in Excel. I worked off of a normalized delta based on the difference between the ratio and the MA of the ratio (varied it some). For all my tests I found that to get the full effect you had to enter on the close and then hold through the next day.

    I didn't feel that to be a viable way to trade (for me) given how some stocks 'move' on the closing print(s). I was concerned the spread could tighten (or narrow) substantially as you get your fill. You'd then be sitting on a overnight with the reason for taking it gone or diminished. Of course, it is a hedged trade so that would help.

    If you waited for the next open you'd often find the spread much narrow (or wider) on the opening prints. Again. all or most of your reward was gone.

    I then moved on to the intraday time-frame with this and other pairs. As I said before, I think the pairs strategy is viable, as attested to by the many folks successfully using it, but I have yet to arrive at a comfortable approach. One thing I've learned (and I have many lessons to go) is I have to be comfortable with the approach. It has to fit my personality.

    I also believe that people are drawn to pairs thinking it removes a great deal of the risk. I've personally seen people trading pairs in such a way they were one COF/PVN away from busting out.

    I, too, would like to hear the quants opinion on this..
    #20     Feb 16, 2003