I just eyeball the ratio chart, looking for a range bound pair ratio. I also look at the slope of MA. You could use some trending indicators like ADX, or calculate the slope of the regression channel. There are many ways to quantify the trendiness. This is not to say trending pairs are not tradable, just not my style. I tend to stick with newsless divergence. You can trade earnings profitably if it's your style
I took another trade (30-10-2013) for the journal and discussion. XEC long @ 106.25 SM short @ 90.00 and here is the chart of the pair.
And another one (30-10-2013), actually two and I've combined them into a three stock pair to disperse the risk. BHP long @ 71.31 BTU short @ 19.89 TCK long @ 27.94 BTU short @ 19.89 I'll post the chart only for the first one, because i can't add two attachments.
Follow-up on this trade - as it looks will close around the flat or marginally positive/negative result. Currently is 0.14% in the money.
I'm starting to think that doing a cointegration test is maybe unnecessary. Your latest batch of trades are from stocks within the energy sector. With a little research on the candidate companies within the same sector you could find stocks that are good for pairs trading. They may even be cointegrated if you ran and analysis on them. I have a spreadsheet with an ADF-based cointegration for testing candidate pairs. I'll run some of your pairs through it at lunch and report what I find.
I have not empirically tested it, but my gut says that cointegration breaks apart as much as correlation does. There will be periods with high cointegration, and there will be periods with no cointegration. If the pair cointegrates constantly it must be something like there once was RIO and RIO.p /the common stock and its mandatory convertible liquid derivative/, but I'm sure that the correlation in this case will be as strong and tradable as the cointegration. Just my idea though, prove me wrong if it not the case.
I know I said I'd post some cointegration results but I never bothered because I couldn't reliably find cointegrations among stock pairs in the same sector. I tried pairs combinations among the stocks recently posted here over various periods of various period lengths. Nothing jumped out at me. I'm sure the actual numerical analysis is correct so clearly there are nuances I'm missing. I've been doing some reading on pairs trading and found some common sense approaches that aren't as fancy (i.e. don't use cointegration)
Correlation is all that matters. -.9> is best,>0.9 works well, too, then backtest. While I can Save you the time, I can also say that correlation when it is greater than -.9 or less than .9 is found in nearly 99.99% of all stocks, So using a more calculating method to determine whether the correlation is negatively high enough or positively high enough is the best way to determine if securities can be profitably pairs traded even before knowing what the back test results show. Older versions of pairs trading software only checked correlation so since cointegration is why emh is still accepted of course you wouldn't find anything but correlation to suffice for a profitable pairs trading method.
This may very well be a one-off attempt to pair trade stocks of companies from different sectors. Results were a gain on the second layer and near break-even for the first layer on a multi-month hold. http://www.elitetrader.com/vb/showthread.php?s=&postid=3916215#post3916215 Although I will keep an open mind, pair trading stocks of the same sector has such rich potential for gains, there is no compelling reason for searching out greener grass.