Disappointing December retail sales. COST overpriced in pair; TGT underpriced in pair. Code: 1/3 1/6 Diff COST 73.45 70.47 -2.98 TGT 60.97 54.77 -6.20 For $10,000 COST TGT Loss Equal # of shares 74.39 shares 74.39 shares -239.55 Volatility weight 80.26 shares 67.32 shares -178.21 Dollar Neutral 68.07 shares 82.01 shares -305.59 Inverse Beta 66.64 shares 83.73 shares -320.54
Afto I dont think you have given the robot a chance. If you had of done some research before hand you would have know that the risk to reward was 1:7 and it had a win rate of 95%. If you were unhappy with this you could have choose not to trade the system . You have completed 10 trades, the backtested results cover 12 years and thousands of trades.
Maybe so but there is a difference between knowing and doing and it just didn't work for me. I'm not saying that it wouldn't for anyone else but its always a question of meshing the trading style with your own personality. In my regular trading I shoot for a minimum 2:1 RR and can handle small losses. To me this was too much like picking up dimes in front of a steamroller. JMO
Got around to looking at the scatter plot for COST/TGT. It doesn't look too cointegrated: Here are some pairs that I like better. SCCO vs BHP: BTI vs MO: DHI vs PHM:
I challenge any of you to make some youtube videos of you guys trading pairs. Please?! . The few on youtube really suck!.
How is scatter plot got to do with CoIntegration? Or is it just correlation? http://en.wikipedia.org/wiki/Scatter_plot http://mste.illinois.edu/courses/ci330ms/youtsey/scatterinfo.html
By definition, two cointegrated markets have a stationary linear spread. In a scatter plot, the linear relationship is usually evident. By itself, this doesn't mean much (everything is tied to inflation, for example.) Consider the following pair which isn't cointegrated, and isn't correlated over long periods: (The last three years, there has been around 80% correlation of price and log(price), but only 15% correlation on return.) The linear structure is what we remove (hedge out) in pair-trading. The residual term, the spread of the scatter plot is what is traded. If the spread of the plot is tight and uniform, such as , there is a consistent relationship (common factor) which continues to hold even at extremes. In convergence pair-trading, we are looking for such a relationship to drive spread convergence. Ideally, according to the literature, the scatter plot would be a cloud of white-noise about the linear relationship. Since we are looking for tradeable pairs, we can tolerate some contrary structure and non-linearity (such as a smile), but without a degree of uniformity, there is no basis to the trade.
Sept/Oct/Nov were all losing months for me. That was a first - to have 3 losing months in a row. December came storming back and I ended Q4 flat, which was also a first for me as I have only had single losing months with each quarter being profitable. I think those kind of draw downs are inevitable from time to time. It hasn't shaken my conviction in the approach. I think the danger when such a phenomena occurs is to change tack and look for something new. Just a hunch - but maybe why Jared turned his attention to FX?