Ok! Gloves off! Correlation, as often experienced in the market, has a temporal lag to it. Not in a million years. We have analyzed Bid/Ask data for over 1000 stocks on 10 years of data. There is no tradable lag! Illusion, my friend! Often when you run macro analysis you find things to be highly correlated in the same time period, yet on the micro scale you will find that one stock is the leader in any movement. This temporal lag is in most cases the result of pair traders, and could even be considered a self fulfilling prophecy of pair trading. Either way, it still presents one way to capture micro anomalies in the market. We have analyzed time intervals from 1 second to 1 week. No tradable time shifts were found! There were occasional anomalies but nothing tradable! On the simplest level, if you are trading a company like MWD, I fail to see how watching and gaining information from MER, GS and LEH is going to be a detriment. More information is always a positive. You can watch anything you like, but only in case of a time sequence your observations are valuable! Meaning if you can not find a case like "If I see $0.1 gain on stock A then stock B will gain $0.2" then you have no additional information therefore your decision is independent on your observations! And because those kind of predicates don't exist due to the market efficiency pair trading is a sexy bullshit and nothing else!
This is ought to be good! An ostensibly naysaying theorist vs. a highly successful day-to-day, in-the-trenches pair trading practitioner/manager/teacher. I say "ostensibly", because this "debate" may just be a clever ruse by the prognosticator to get a 3-day pair trading seminar for free...!
Well sure I can understand the comments, and it is clear that you are correct. A couple of things occur to me however. First, the difference in volatility between the two stocks and the spread is incremental in my view. Neither stock is highly volatile historically, however the potential to see significant price shock is always present. I think my own view of the spread process is different than yours. For me, I am looking for correllation first, and for evidence of increased volatility beyond what is customary for the pair. That is my signal to bet on reversion back to a less volatile state. I understand that volatility = risk, and I look to get paid by evaluating my potential profit in light of practical considerations like position size and risk managment (possible hedging). If I cannot protect myself adequately I wont trade the spread. As for efficiency, again believe the probability of profit is greater with the spread, so the price I must be willing to a pay is a reduced magnitude of profit. Also I have to look at other opportunities that might occur during this temporal window. If for instance, I had invested in a residential property during this month instead of betting on Royal Dutch, I would have made much better profit in absolute dollars, but I would have been exposed to much greater risk of collapse of the real estate market as well with no opportunity to hedge that risk. Lefty
If you are dealing with 50/50 probability then any product of any two variables with this probability will have the same (50/50) probability! More so, "predictability" as you were saying has nothing to do with pair trading it has everything to do with probabilities of individual stocks only as any product of them is their derivative.
There are by far better ways to achieve the same purpose. One of them is collar: - Long underlying, short call and long put - for your long bias and Short underlying, short put and long call for your short bias. You can design any protection level you want and achieve by far better returns. Synthetics are much more powerful and effective hedging tools than pairs. Also, I can show you a synthetic for any combination of underlying you want! And in every case my structures will be more efficient!
The market is slow, I am making money, my automated programs are running and I get to talk about trading... life is great! Not in a million years. We have analyzed Bid/Ask data for over 1000 stocks on 10 years of data. There is no tradable lag! Illusion, my friend! Tradable is a matter of transaction cost and execution skills on the micro level. It is impossible for two stocks to print in an identical time path, there must be some form of temporal lag. The question comes down to, as you have shown, is if it is tradeable. The market structure of the NYSE is beneficial for this with non automated trading programs as it creates inefficiencies and the ability to get stock if you know how the system works. People are trading for many different reasons and often you will find bids or offers to be able to hit from people who are not trading the correlation. We have analyzed time intervals from 1 second to 1 week. No tradable time shifts were found! There were occasional anomalies but nothing tradable! Generally most people look at spreads and argue that they look amazing to trade with their mean reverting properties. It is the execution of the extreme prints that always is the tricky part. Hedged trading has been around since the inception of the market, the only difference with pair traders is that we use another stock in place of an option or derivative to do it. There is a time and a place for each. There is no doubt that you give up some of your returns to be in a hedged position, but some traders are willing to do this for the benefits of consistency and predictability. You can watch anything you like, but only in case of a time sequence your observations are valuable! Meaning if you can not find a case like "If I see $0.1 gain on stock A then stock B will gain $0.2" then you have no additional information therefore your decision is independent on your observations! And because those kind of predicates don't exist due to the market efficiency pair trading is a sexy bullshit and nothing else! Correlation implies that "If you see a gain on stock A then you will see a gain in stock B". That is the information you have!!! The question is just can you execute it. For longer time periods more things then correlation get involved: like fundamentals, news events, reasons why the correlation has not held yet, but should in the future. Markets are not naturally efficient, anomalies are all brought back into line by traders. Often times you may be forcing stocks to be correlated by pair trading them and bringing about the market efficiencies. This is something that is only done if it is profitable for the people doing it. There are many ways for people to make money in the market, it is about finding the method that suits your personality. If you can handle being one sided in the market and the risk that represents, while making money consistently then don't come to our seminar! Keep doing what you are doing! It is an alternative method that fits for some people and allows them to make money. Cheers, Darren PS Keep it up!!!
Gain of information from covariance and the predictability of covariance. In essance that is what we are arguing.
They're "long term" traders, man. They want you to get used to the concept. Plus they may want to give you a little time to save the $1500 entry fee. Are any showgirls included? Coinz
Covariance is a statistical measure of the variance of two random variables that are observed or measured in the same mean time period. This measure is equal to the product of the deviations of corresponding values of the two variables from their respective means and does not contain any "NEW" information.
Correlation presumes no time shift! Correlation is an integral of multiplied changes of two variable occurred at the same time.