PairFinder

PairFinder is an app to find cointegrated stocks to pair trade. When two stocks are cointegrated, it means they move in the same direction at the same time. When one stock goes up, the other goes up too, and when one goes down, the other goes down. A regression is calculated on the historical stock prices, and an equation is determined for the stocks. Then one stock is calculated as a multiple of another, plus a constant value, and a 'random' error variable. The random error is calculated at each point by subtracting the expected value of the second stock from its actual value. This random value is then statistically analyzed to determine if it is a 'stationary', 'constant' random variable, or if it drifts and is a 'random walk'. If the probability is > 95%, then it is (most likely) a constant random value, and the two stocks are cointegrated. If the probability is < 95%, then the random variable drifts, and the two stocks are not cointegrated. The cointegration test is an Engle-Granger test with Augmented Dickey-Fuller (ADF) test for 'unit roots' in the random value 'residuals'.


Pair trading is a method used by advanced hedge funds to take advantage of arbitrage opportunities. When stocks A and B are cointegrated, they will normally be the same 'distance' apart. When they diverge or converge, a trading opportunity presents itself for profit. If stocks A and B diverge, A goes up while stock B goes down. A trade is made to sell (short) stock A, and buy (long) stock B. When A and B converge and come back together, then stock A is bought back (buy to cover) and stock B is sold. Pair trading is a low risk, low capital exposure trading strategy. Trade are open for a few days or a few weeks, and the capital required is minimal due to selling and buying stocks at the same time.

PairFinder requires a subscription of $0.99/m. Background data must be enabled for the license server.

Invalid symbols are shown as 0% probability, as are symbols that are missing price history for a given time (ex. FB/MSFT for 06/2011 - 06/2012).

In the future:
* Ticker symbol autocomplete and company name autocomplete.
* Global symbols. Currently only US exchanges available.
* Industry and sector selections, to find pairs within entire industries.
* PairTrader, to send pair trading signal notifications when a trade opportunity arises for selected pairs.
* Regression coefficients to determine price relationships.
* Portfolio and pair allocation based on available capital.

PairFinder was created using Java for the app, and Perl and C for the server-side computation. A modified Gretl econometrics library was used, as well as GD library.