Tuesday, December 7, 2010

Monthly update

Back after a two week break. Did some early season riding in Maine. Have not had the time to update the blogs.  Eventually I plan to move this trade signals platform over to collective2.com. I spent significant amount of time in enhancing the trading process. I will post additional results by the end of this month. I promise there will be some interesting findings.  It will give a better view into some of the industries, the return process, and back testing results of the portfolio.  The system is flexible enough to run for short term data as well as long term data. The only problem is the time required to run a 20 year simulation. To give an example, the commercial banks industry has 420 publicly traded names in the US. If we run co-integration test for all pairs in this industry, we are talking of 90 thousand computations for one day. When we run the process for 20 years the problem grows accordingly. We can add additional filters to reduce the number computations. Since I run four threads (because I use a quad machine) the run time is reduced signifincatly but still we are talking about a week to run the trading system for the entire US stock market (60+ industries) for a 20 year time window.


Now to the current portfolio. The pairs positions are being reduced mainly in anticipation of new findings. Better safe than sorry. The gross leverage in the portfolio is now about 130%, net of +50%.
I still don't have a complete picture of November performance.  For most part the month ended flat driven by losses in Google. This had to do with the inability to roll the hedges (on time spread) on black Friday (was on the slopes riding on first snow). No Groupon, and now google seems to be on the come back.  Closed several positions, but didn't have the time to update the posts. Will do a more detailed posting soon.

For now, GDP/TLM has been closed. KDN/WAB has been closed, a few opportunistic shorts on NFLX, and AAPL during last week worked well. The strangle on NFLX has been closed as well. The strangle could have been more profitable if I had used a balanced approach. The slight negative bias on NFLX worked against the portfolio.  January 600 Call on GOOG has been opened at an entry price of 13.2.

1 comment:

  1. If you have a NVIDIA graphics card in your Quad-core, try using CUDA which is just an extension of C and you can have a lot of threads running simultaneously which should speed up your simulation time considerably

    ReplyDelete