Need help with a Pair Trade XOM - CVX

Discussion in 'Strategy Development' started by yobo, May 1, 2008.

  1. yobo


    Hey guys thought I would try to get some feed back on the xomcvx pair I have going as I am new to pairs.

    The correlation is about 94.5% over a 252 day period.

    The "mean" is 1.005 for 252 days.

    The Standard deviation for 252 days is .0175

    Yesterday the pair traded at a spread of .968 and two standard deviations away from the norm is .965 so I entered the trade long an equal dollar amount of XOM and short CVX as CVX was the best performer.

    Today, XOM reported earnings and is getting killed and CVX is also down as well but now, the spread .94 or aproximately 3 standard deviations away.

    My question is this, do I add another layer onto the trade and go long XOM and short CVX?

    In an ideal world, I want to be long CVX as that has been the better performer over time, but since the pair has deviated so much from the mean with XOM getting whacked, mean reversion suggests the pair will come back so adding another layer seems to make sense.

    Hoping to get some input from experienced pair traders on what to do. Thanks in advance.

  2. I cannot really advise you on this particular position - other than to point out the obvious. XOM is down for a reason. It missed the earnings expectations.

    To prudently add more layers, you would have to hope, uh, ... "know" ... that CVX will have a significant miss tomorrow as well. Only you know what you think you know and what you think you don't know.

    Going forward, I would advise you to not be so married to the correlation / standard deviation stuff - or - in other words, put it into perspective.

    1. 2 and sometimes 3 standard deviation moves happen. Do not confuse "unusual" with "impossible."

    2. They (2 and 3 standard dev. moves) tend to happen around earnings. Like now.

    3. As you have pointed out - the stocks are very highly correlated. The 3 yr. correlation is +.94. However, you would be very naive to think that this type of price action has not occurred multiple times over the past 3 yrs. In other words, a high correlation does not insulate you from this type of price action.

    4. Even high correlation spreads can trend. XOM missed today. If CVX beats tomorrow, the chances that XOM outperforms CVX over the next few weeks or months go a bit lower.

    As a newbie and maybe even ultimately as an experienced spreader - you may want to consider strategies where you sit out earnings.

    Just a thought. Good luck.
  3. yobo


    THanks for the response. Was wondering if anyone would offer feedback.

    Here's a clip from an academic paper regarding pair trades and returns:

    Abstract: This paper examines the impact of accounting information events (i.e., earnings announcements and analysts’ earnings forecasts) on the profitability of a pairs trading strategy proposed by Gatev et al. (2006). Using a portfolio of U.S. stock pairs between 1981 and 2006, we find that pairs trades are frequently triggered around accounting information events.

    "More importantly, we find that pairs positions opened after accounting events are significantly less profitable than pairs positions opened in non-event periods. Furthermore, we find that incremental excess returns can be achieved by delaying the closing of a pairs position until after accounting information events."

    Overall, our results suggest that drift in stock prices following earnings announcements and analysts’ earnings forecasts is a significant factor affecting the profitability of pairs trades.

    Here's the full link:

    The study actually seems to indicate that its best to open the pair before an event such as earnings and then to close out afterwords. This is opposite of what you are suggesting?

    With that said, I entered the trade with 2 STD deviations away from the mean and the pair is now trading at 3 deviations. I've opted to enter another layer at 3 STD's. ANd will again at 4 STD's

    I double checked the data and over the last year the pair only traded at this spread differential a few times based on closing prices. So todays spread seems to an anomoly and perhaps a real opportunity for some great profit.

    Thanks for your feedback. The only real concern I have is that based on fundamental valuations I'd prefer to be long cvx and short xom, but the spread indicated otherwise so I went for it.

    Live and learn.
  4. yobo


    Here are the layers I have put on...

    Layer#1: Long XOM 92.17, Short CVX 95.19.
    pair ratio = .968 or approximately 2 std dev...

    Layer#2 Long XOM 90.26, short CVX 94.46.
    pair ratio = .955 or approximately 2.5 std dev...

    Layer#3 long XOM 89.21, short CVX 94.41.
    pair ratio = .944 or aprroximately 3 std dev...

    I post this in case others are interested in pair trading and want to see how one actually works out.

    My trading plan calls for adding layers at each incremental std. deviation. As I type this, the trend has changed and the spread is coming in. My last trade, the spread was .944, it is now .951. The last layer is now profitable.

    I plan to hold the layers through CVX's earnings call on Friday and I suspect, sometime either tomorrow or next week, the pair will revert back to its mean of 1.00. I will probably exit the trade at .99ish.

    Stay tuned.
  5. 1) Do not let losers get out of control
    2) Correlations are random and variable. They are not "certain".
    3) Things can diverge farther from the mean than you expect.
    4) The mean is not a stationary target.
    5) Trade and learn!
  6. yobo


    Trade and learn...well that is what I am doing. I am sticking with the pair because even the though the spread has been increasing, the positive correlation remains in tact. Positive correlation reffering to the fact that if one stock trades down, the other trades down and vice versa.

    The widening spread creates opportunity for a very profitable trade. Both are moving down just at different speeds. I would only stop out of the pair if the correlation changed and turned negative. An example here would be XOM heads lower and CVX heads higher. This would be negative correlation.

    Anyway, this is an experiment for me. Positions are small. My time frame for the trade to work is longer than just today.

    Thanks for the words. Do you actively trade pairs?
  7. DonKee


    There's a reason why most "pair traders" have anywhere between 20 - 100 different pairs on at any one time.

    There's also a reason why many pair traders, trade legs in and out throughout the day.

    There's a whole lot more to successful pairs trading than just putting them on at various deviations from the mean and taking them off when/if they ever revert back to a level they seemed to trade at in recent history.

    Take a look at GE/HON or RIG/DO over the past 5 years just to name a few.

    What was once normal, is now abnormal.
  8. First - you are doing well by experimenting and keeping the size small. Very commendable.

    However, in the true spirit of experimentation - you must ask yourself "Is what I am doing right - and if so why do I think I am right?" as opposed to "This must be right because I am doing it"

    You seem to be doing more of the former vs. the latter - although I am not sure I agree with some of your key conclusions.

    Re: The study - I think I am familiar with it. If not this particular one, then others that have similar conclusions.

    What is missing in your post is your criteria for being long XOM vs. CVX.

    And, also, the criteria the study uses to pick the long vs. the short (I won't have time to look at the link and read for the specific criteria used in the study until after the close).

    Many of these studies identify some fundamentally "expensive stock" via P/E or some other measure(s) vs. a "cheap" stock via the same metric or set of metrics.

    The theory is that the high fliers fall harder when they miss vs. the rallies they tend to have when they beat. And the downtrodden rally harder when they beat vs. when they have missed for the third time in a row.

    So to capture that - you would have to hold through earnings.

    I have a lot of respect for academic studies and find them helpful. But they have a difficult time capturing the nuances of intraday and relatively high frequency short-term trading. Many of these studies are kind of "buy(or short) and hold" from one quarter to the next based on the criteria being studied.

    From a pure price action perspective, there are tons of opportunities for trading between the quarterly releases. It is difficult for these studies to capture that dynamic. Also, the nuance of adding layers, how wide or how narrow, for how much profit, etc. - is rarely, if ever, addressed.

    To your other question, yes I trade pairs - but I am not a big fan of intra-industry mean reversion. To me, picking the long vs. the short with XOM vs. CVX is like being paid to tell which Olsen twin is Ashley or Kate from a distance. Possible? Sure. But identifying Hulk Hogan vs. Paula Abdul from a distance is a lot easier, IMO.

    So I specifically look for a security that will that will outperform another security.

    Back to your study (or at least the ones that I am familiar with) - I think the difference in the "cheap vs. expensive" metrics were pretty significant between the longs and the shorts. I haven't looked at the fundamentals of XOM or CVX - but I would be surprised if they are that radically different. So the "edge" of being short a high flyer vs. being long a downtrodden stock may not be there in this case.

    Anyway, there are 1,000 ways to skin the pair trading cat. As with trading in general, the key is to find an approach that woks well with your personality and your resources.

    Good Luck.
  9. yobo


    Thanks for that response. Very insightful and I think you are right on about your assumptions concerning the academic paper.

    THe one flaw I have with being long xom and short CVX is that fundamentally on a valuation basis, the data indicates be long CVX as that is the better value play.

    However on a 1 year trading history, the stats indicate the other, long xom and short cvx based on the statistical arbitrage opportunity.

    Tomorrow, CVX reports and I'll be watching for a negative correlation. In otherwords, CVX goes up and xom goes down versus both going down at different speeds as we are seeing today.

    To me that would be a strong signal to exit the trade as correlation begins to show signs of making a sea change and the time value of money changes significantly getting pushed way out as the time for the pair to revert back to the mean becomes longer.

    Thanks again. This has been very helpful. The one thing I am learning is that perhaps to be real succesful at pairs is to automate the process so one can trade the intraday easily and manage a lot of different pairs at the same time.
  10. great thread thanks guys...
    #10     May 1, 2008