Most quantitative plays are also impossible to execute except on vast scales. Even if they are identified by retail players, they cannot be exploited because the margins are thin and the time-windows are very small.
Very few quantitative plays seem to hold up for a reasonable length of time and offer sufficient returns for retail players to contemplate. One of those few is the coupling or pair-strategy.
The couple involves selling one stock and buying another stock. Actually it usually involves trading futures on two stock underlyings. Here's the process:
Find two stocks that trade in such a manner that the price difference between the two remains nearly constant. Every time the price differential is larger than normal, or less than normal, assume that it will correct back to the mean.
If the difference has increased, sell the higher-priced stock and buy the lower-priced stock. If the difference has decreased, sell the lower, buy the higher. Reverse when the mean reversion occurs.
A pair strategy is not difficult to understand or identify. Execution can be tricky but not impossibly so. It works in many market conditions though some conditions throw up more pairs than others. The returns can be large maybe even 10-15 per cent in a 3-5 session period. The risks are not very high for a disciplined trader.
There are some interesting assumptions and technical points involved. The first is that the mean reversion the trader is betting on is only likely to work where the two stocks maintain the price difference by moving in close tandem.
It's not too difficult to identify stocks, which have a historic pattern of close correlation. They usually come from the same industries. A word of warning though, they may not be true sometimes.
For instance, a glance at Mahindra and Mahindra versus Hero Honda will show that these two aren't an ideal pair. They have frequently moved in opposite directions. At various times, the hitherto more expensive stock has been overtaken by the less expensive one.
Reasonably stable pairs that seem worth examining include Infosys versus Wipro, HPCL versus BPCL, Ranbaxy versus Dr Reddys, or Siemens versus ABB. There are many more possible pairs. Just remember there should be a price difference and a strong correlation in movement.
Decoding the statistical details is not so difficult. You should be able to figure out if the distribution of price differences for any two given stocks is normal or near-to-normal in its pattern.
The next tricky detail is working out the futures lots sizes and ensuring that both legs of the trades are approximately the same size. In an ideal pair, the two legs of the transaction should be equal in cost.
This may require a scenario where for example, the trader buys two lots of stock A for every three lots of stock B sold. Sometimes the pair is otherwise ideal but the futures lot sizes make it cumbersome to execute trades.
The third thing is setting tight stops on the transaction. There are times when a history of mean reversion breaks up. You don't want to be caught with an open position in that situation.
Again, the recourse to statistical methods and the use of standard deviation as a risk-control tool helps. You know how much you hope to gain when you enter. So you can set a cut-off loss limit based on your risk-reward profile.
The ideal conditions are found during flat or somewhat bearish markets. You don't want a runaway trend developing because that makes the mean reversion less likely. In a highly bullish or highly bearish market, there are better trades available in any case.
One interesting situation is during periods when results are declared. Say, you find two highly correlated stocks that generally maintain constant differential. One declares results ahead of the other.
That causes a momentary increase or decrease in the difference and gives a clear window inside which the reversion is likely to occur. Obviously the current market conditions and the results season makes it a good time to hunt for pairs.
