A quick side question about another type of algo, the pairs trading algo.
This algo I think looks for correlation between two stocks in the same sector and if one stock, stock A starts going up the algo will immediately start buying the other stock, stock B that is correlated to stock A.
However since the volume that the algo can bring to bear on the market is so great, does it not start to create artificial correlation?
Take the fact that Stock A and Stock B have been historically correlated but an event in Stock A's corporate structure warrants a better outlook on Stock A causing it to rise. Now if the algo's were not there, Stock B would rise but by a lesser degree since the event only revolved around stock A.
But I was reading an article that claims because of algo's the markets have become much more correlated with each other. Algo's lartificially enhance the correlation just based on their past statistical correlation on Stock A and B without taking into account the real fundamental value of Stock B and whether or not it deserves a higher price.
basically the problem with just trading based on statistical correlation destroys the old Market portfolio theory of Markovitz that many investors use to protect themselves by diversifying their assets. The idea that assets that are non correlated will protect them from wilder swings in the market
But since algo's have so much volume behind them they can push prices of Stock B only because Stock A went up and they were historically correlated.
In the past Stock B would go up just on the amount the investor buying stock B, thinks this structural chance in Stock A would effect Stock B. The demand for Stock B would be real. Of course there would be human traders that would buy stock B because they know they have been historically correlated.
But with the amount of buying power some firms employing these algo's, they can make stock B more correlated than they have ever been.
So algo's that trade on correlations have the ability as a whole to make the market more correlated and unidirectional not based on real value and real interaction of different ideas of where price parity should exist but just based on statistical data.
So now knowing this I can exploit this, if I notice stock A rising after a major news event, I know the algo's are going to buy up stock B and I should get long stock B.
Maybe I have my logic backwards, just curious as to what you guys think about this type.
This article makes some mention of this
http://www.risk.net/energy-risk/feature/2136141/algorithmic-trading-energy-markets
In paragraphs 9 and 10