Hello, I have discovered that numerous equities pair traders use the ratio of the 2 stocks to model and trigger their trades. On my end, even if I have never really traded stat arb( apart from a futures automated system in '07 ), I would use difference or more exactly beta adjusted difference instead. For example, if ABC @ 92 and DEF @ 42 are super cointegrated and ABC has an ATR of 1.5 times the one of DEF, I would naturally trade ABC-1.5*DEF but never ABC/DEF or even ABC/1.5*DEF. Am I wrong here? Who does like me?
With stocks the theory seems to be to neutralize market and sector risk by volatility matching the pair. Some other ratio may yield a better line for your signalling system to trade but you're not neutralizing market and sector . Large volatilities differences between legs of the pair applied to a simple ratio x/y leave you trading the more volatile leg somewhat outright, dampened a bit by the other leg. Two most mainstream methods are relative performance divergence(alpha through trend) and convergence/MR ,basic stat arb . Got to get direction right either way in my world. That's my take on it. Importantly, the theory goes that you have to select pairs that make sense to neutralize market and sector and avoid coincidental/accidental cointegration.
An other basic question: Are you all using complex models, backtests and "hard" metrics to trigger a trade? I am trying to understand all the metrics on PairTradingLab...I don't even know what a Z-Score is. I thought I would use a more basic approach. Spend most of my time to find a few of those "heartbeat" spread charts and then place trades visually...
You want to construct your chart to match the model you are trading. If you are trading relative strength where you are expecting one stock to move more than the other in the pair, then a ratio is appropriate. If you are trading stat arb where you are expecting the stock prices to converge to their mean value, then a difference is appropriate.
Let me restate my last post more clearly and accurately (I'm unable to edit it): You want to construct your chart to match the model you are trading. If you are trading a relative strength model where you are expecting one stock to move more than the other, then a ratio chart is appropriate. Using a ratio chart assumes you are dollar neutral. If you are trading a stat arb model where you are expecting the difference (spread) between the stock prices to converge to their mean value, then a difference chart is appropriate. You can weight for volatility or whatever you feel is appropriate. The term "pair trading" usually refers to stat arb. The term "spread trading" can refer to either stat arb or relative strength trading so you have to be careful to define the context.
Is it true that more and more traders are switching to arelative strength model compared to stat arb? Is it more suitable for a low vol environment?
Stat Arb (aka, convergence trading) is challenging because there are so few trading instruments that are cointegrated with consistency. You can apply a Relative Strength model (aka, divergence, trend, or swing trading) to any number of instruments, so there is no difficulty in finding markets to trade. The challenge here is developing a model for predicting Relative Strength. I don't know why a low vol or high vol environment would have anything to do with the effectiveness of either of these trading approaches. You might check with ET member bone (who teaches spread trading) to get his POV on what most spread traders are practicing these days.