In my case I'm looking at the cor. of the last 365 days. The two stocks need to be in the same industry.
Thanks for the reply. So how often do you find that a relationship breaks down after being correlated for 1 year? I am guessing you use stops to protect yourself against this.
Some pairs disappear from my radar from time to time because of that. But that does happen very rarely from one day to another. I'm very selective when it comes to pick pairs for my trading. I use time stops, but no regular stops.
Thanks Colonelangus. Even a blind squirrel finds a nut now and then. Nice to have a decent day in a tape that seems to reward momentum players far too much, but hey, maybe that's just me. Glad you made a little, too. I took most of it off because I am famous for exiting winners too soon, and why break the streak now?
I like this idea. 77% correlation and appears to trade in a pretty well-defined range. If it doesn't gap in the morning, I may look to put some on for a trade. Thanks for the idea.
I personally use 500 trading days (2 years) for correlation. I also use that time frame when calculating standard deviation of the differential and ratio, which I use for longer term plays. I just can't help myself when I see a nicely correlated pair that looks ugly, but is 2.5+ std deviations from the mean on that time frame. I'll put it on in most cases, knowing that I'll probably have to take some pain in the near term. Probably best to see it turn first to put some on the sheets. But if I position size and diversify correctly, it will work out fine in the vast majority of cases if one can just close their eyes and look out 3-6 months. KIM/CLI is a good example of this.
It's great to get feedback, thanks! In general, do you guys have a set of pairs that you are constantly watching, or do you scan over all stocks looking for new correlations? Do you guys use commercial software or have you developed your own?
Most guys here have their fixed pairs I think. I scan all the combinations of about 1400 stocks everyday and make the numbers (now cointegration, not correlation). Most use Pair Trade Finder dot com. They seem to like it. I and at least another have our own soft. It's a lot of pain to develop your own, unless your are very experienced in coding. But it's flexbile.
The name "Pair Trade Finder" suggests that they can also scan all combinations everyday, just as you do. It makes good sense to do it so I guess that everyone would be doing this. Or would I be wrong in saying that?
I think you have to input your own pairs in the software. In practice, if you review your list every once in a while there is no need to scan all the planet everyday. I do cuz I'm too lazy to look for good pairs and look at only them on a daily basis. I also get a lot of shit in my scan everyday. But then the question remain, how do I find good pairs? I dont know what others did. This can be a lot of work. What I would do is select stocks that you want to trade based on some criteria (price, average daily volume, ...) and then pair all the ones that are in the same industry. This can give hundreds if not thousands of combinations. And then look at them one by one. I think the soft has a backtest feat., that could be a metric.