Page 1 of 1

Correlation and pairs trading

Posted: August 4th, 2011, 5:27 pm
by chicagokid
My group does equity pairs trading using a stat arb strategy. We are however, intraday only. Without delving into details, we have 6 strategies, which differ with respect to various parameters. Conventional wisdom dictates that if stocks A and B are correlated, a mean reversion strategy should be profitable on a fairly consistent basis. However, over the past 2 weeks or so we've been doing very well in agricultural pairs (CF-POT, POT-MOS, CF-AGU, CF-MON, etc.,) and getting crushed in industrials and natural gas pairs. All of these pairs have been moving in the same direction, especially since the market has been taking a brutal beating, and the S&P 500 stocks are virtually all moving down, but I can't figure out why pairs in those sectors are doing so much worse than agriculture. I don't quite know how to get to the bottom of this, so any advice would be greatly appreciated. Doing a cointegration test would probably be useless since we're intraday only.

Correlation and pairs trading

Posted: December 12th, 2011, 7:31 pm
by prfj
Are you sure about "Conventional wisdom dictates that if stocks A and B are correlated, a mean reversion strategy should be profitable on a fairly consistent basis"? Really sure?

Correlation and pairs trading

Posted: December 12th, 2011, 8:23 pm
by acastaldo
*deleted*

Correlation and pairs trading

Posted: December 12th, 2011, 8:25 pm
by HOOK
QuoteOriginally posted by: chicagokidConventional wisdom dictates that if stocks A and B are correlated, a mean reversion strategy should be profitable on a fairly consistent basis.Correlation does not imply mean reversion. They must cointegrate to garantee that.

Correlation and pairs trading

Posted: December 13th, 2011, 9:51 am
by Marine
As everyone has already mentioned, correlation does not imply mean reversion! Also a pair does not need to be cointegrated for mean reversion to exist.Cointegration can exist intraday, just because you haven't found it to be true does not mean it does not exist.And finally just because you think something is cointegrated or a mean reverting process does not necessary mean that you will make money from trading it.This happens all the time ... you trade a pair which makes money from MR and then one day it stops. The way you manage you risk and determine when to enter/exit these pairs will determine your survival.

Correlation and pairs trading

Posted: December 13th, 2011, 10:34 am
by DevonFangs
QuoteOriginally posted by: chicagokidDoing a cointegration test would probably be useless since we're intraday only.I don't know the first thing about this, but am interested. Can somebody please elaborate on this.

Correlation and pairs trading

Posted: December 13th, 2011, 11:28 am
by ktang
QuoteOriginally posted by: chicagokidI don't quite know how to get to the bottom of this, so any advice would be greatly appreciated. Doing a cointegration test would probably be useless since we're intraday only.I also dont understand this point. What you can do is to calculate the value function under a stochastic control approach. This will also tell wheter you should trade or not. I mean just becuase the spread is mean reverting does not mean you should trade.

Correlation and pairs trading

Posted: December 13th, 2011, 12:51 pm
by PavelG
QuoteOriginally posted by: HOOKQuoteOriginally posted by: chicagokidConventional wisdom dictates that if stocks A and B are correlated, a mean reversion strategy should be profitable on a fairly consistent basis.Correlation does not imply mean reversion. They must cointegrate to garantee that.Yes, correlation does not imply mean reversion from theretical point of view, but from practical point of viem, it is enough correlation. Can you give example from market?