Note: Anyone that wishes to see an example spreadsheet using calculations to trade pairs please email me with the subject headline “Differential DV2.”  My email address is dmvaradi@gmail.com  Furthermore, a comparison of the DV2 versus other methods will be compiled for distribution very soon.

There are many ways to trade pairs: 1) Bollinger bands 2) stochastics 3) RSI and 4) z-scores to name a few. In my opinion, the DV2 (a specific setting of the proprietary DVO which is a short-term ocscillator) and the DVI (intermediate oscillator) are vastly superior. This is not just an opinion it is borne out by internal research.

The returns regardless of the parameter settings when compared appropriately to each of the aforementioned alternatives are substantially higher with nearly identical standard deviation, lower drawdowns, less time in drawdowns, and smaller maximum losing trades. Note that the same results are found when comparing DV indicators on individual stocks or ETFs for swing trading.

The question is why? This was partially addressed in Part1 but the reality is that standard indicators each suffer from certain disadvantages. I would first point out that Bollinger Bands, and z-scores can be grouped together, and so can stochastics and the venerable RSI. Bollinger Bands and z-scores have one primary achilles heel—they rely on the assumption of a normal distribution. Without getting into a heavy discussion of the flaws in the assumption of normality in security returns, suffice to say extreme events are far more probable than what the bell curve or lognormal bell curve would indicate. Furthermore, the actual distribution can also cluster heavily around certain areas, and this is unique to every stock or ETF. As a consequence, both shallow levels such as above/below the mean or extreme levels can be misrepresented leading to poor signals. RSI and stochastics are relative indicators in the sense that a mathematical scale is used to determine readings relative to historical data (as in the case of RSI) and maximum and minimum values as in the case of stochastics. This presents a different set of problems as stochastics for example spend most of their time at extremes, while extreme RSI readings do not neccessarily imply that a certain overbought/oversold reading is actually rare or extreme based on the current environment (like the last week or so).

Additionally all of these indicators fail to take into account relevant intraday data and scale for it in a meaningful way. It is intuitively obvious that intraday highs and lows represent meaningful support and resistance levels. Additionally extreme stretch is captured more accurately by examining the pattern of multi-day price behaviour. In the case of pairs, the very best opportunities occur intraday when volatility creates tremendous discrepancies. This volatility needs to captured adequately to understand when a pair is likely to revert to the mean.

The DV indicators use the “empirical distribution” which is a fancy way of saying that they look at the actual historical data of the stock, pair, or ETF. Capturing all of the strange nuances that are unique to each. Furthermore, this distribution is adaptive because it constantly rolls forward updating the distribution with fresh data. Of course these indicators also utilitze intraday data, which makes them even more effective.

Pair trading and arbitrage is a competitive game, involving small returns and fixed costs in the way of commissions. Any edge that can be extracted can mean the difference between survival and failure, especially when margins are driven down by competition. Using simple, but intelligent refinements that are unique like the DV2 give you an edge that your competitors do not have.

In the next post i will provide specific ideas for different types of arbitrage opportunities and several different strategy ideas (no more gratuitous plugging of my own indicators!).