Every clients of our custom reports have their own strategies and criteria. They tell us their specific requirement and we can generate a special report contains pairs fulfill the criteria. Some of them are asking about what a good criteria is. I have to say there are no good criteria. Everyone has his own investment style and no single strategy suits for everyone. But there are still some ways to approach profitable criteria. My opinion might not be absolutely right, but, after discussed with many pair traders, the following is some clues to a good criteria:
- Trading volume: the larger provides better liquidity and less transaction cost, which is key to good result of arbitrage and pair trading.
- Headlines and SEC filing: prevent stocks with recent headlines or SEC filing especially those quarterly or annual reports just announced. That might alter the NORMAL relationship and the historical price is nothing for trading.
- Distance to regression line: the larger might not be better. Most farest pairs are driven by headlines and SEC filing, the distance of a good pair is usually between 1.5 to 2.5 standard deviation.
- Same industries: same industries is another not so important factor because some pairs not in the same industry have high correlation and keep good relationship. If you can find pairs not in the same industries, your arbitrage opportunity should be much higher than other traders.
- Correlation: it might be strange that I list correlation below some other factors. Many textbooks told us correlation is key for pair trades. I don't think a high correlation is a must to do the pair trade. A perfect correlated pair is absolutely untradable because these two instruments never separate apart. A highly correlated pair is tradable but it usually implies tiny spread even at their historical record. A reasonable range of correlation should be defined in criteria. 0.85 to 0.98 should be considered under different situations.
