Gekkoquant.com Website Review

QUANTITATIVE TRADING, STATISTICAL ARBITRAGE, MACHINE LEARNING AND BINARY OPTIONS. The term statistical arbitrage (stat-arb) encompas*es a wide variety of investment strategies that typically aim to exploit a statistical equilibrium relationship between two or more securities. The general principal is that any divergence from the equilibrium is a temporary effect and that bets should be placed on the process reverting to it’s equilibrium. desirable, however at some point the divergence will grow so large that one has to concede that the equilibrium relationship no longer relationships and to as*es the duration of any observed equilibrium out of sample. This post will investigate the power of the statistical tests in relation to pairs-trading for the following statistical tests ADF, The general principal is that for two stocks and they form a stationary, and by definition, mean reverting pair if the following If is between and then and are co-integrated, is the coefficient of mean reversion. A statistical test must be performed to check if , this is known as a unit root test. If the series contains a unit root it isn’t suitable for pairs trading. There are multiple unit root tests, each running a different test on the residual process. One might be tempted to estimate the AR(1) residual model and check for using the conventional linear regression method calculating the standard t-ratio. However it was shown by Di*ky As with every model there are trade off when determining the training window size, too long a window and the model may contain irrelevant data and be slow to adjust to recent events, too short a window and the model only responds to recent events and forgets about past events tends to 1 from below, for 250 data points with the barrage of co-integration tests only detect co-integration less Intuitively this makes sense, the slower the process is to revert the more data points will be needed to see the reversion. It is somewhat undesirable that the power of the unit root tests vary depending upon the properties of the underlying process, however it is not required