Jan 31 2009: January 2009, my best month ever
I finally get much needed help Our new system January stats
I finally get much needed help
January 2009 was the first full month where we applied our new investing strategy in real time. When I'm saying "we", it is not in the royal "we" sense.
As I wrote a month ago, I'm now grateful to have help. My old time friend, Adi, has moved to the Bay Area to take a position at Google. His impressive programming and Math skills have made a great difference.
I feel Adi is the perfect partner for me; he's a great pleasure to program and brainstorm with, we complement each other while having no demands or particular great expectations of each other. We have an agreed upon simple goal: to develop a personal investment system which "maximizes risk-adjusted returns and minimizes tedious work."
Neither me, nor Adi, is the boss, we work in a very informal arrangement, we are both very motivated to make the goal happen, and we have no schedules or deadlines imposed on us from the outside. We keep making progress every weekend, and are making it a habit to surprise each other with new ideas and improvements to the system when least expected.
From very early on, I felt confident that with Adi, I could take the endeavor to the next level, and that's exactly what happened.
One thing we've done, is to make our code development distributed. Since we like to work on different parts of the code, and since there are so many items on our
TODOlist, we've put all the old code I've originally written in the past several months, using the 'R' programming language into a common repository. By some coincidence, about a month after we started spending weekends together on our growing R code, The New York Times published a pretty decent article about R, authored by Ashlee Vance and titled "Data Analysts Captivated by R's Power" in its Business Computing section.
Adi suggested we use git which was originally written by Linus Torvalds, and Junio Hamano. This was a great choice which has been speeding up our parallel development.
Our new systemWe feel our new system is too valuable and unique to be detailed in full on the web. I can say a few things however which should be pretty obvious to anyone following me for a while.
- The system is opportunistic and uses both long and short positions. As such it cannot be implemented on Marketocracy. Marketocracy has too many issues which make it impossible to emulate our actual strategies on it, as I detailed in previous articles. I'm planning to start using a new third-party auditing system soon.
- The system may use leverage (up to 3x per single position) wherever the estimated upside chances clearly exceed the downside chances. 2009-03-03: clarification: This shouldn't be confused with excessive risk-taking - we are pretty conservative and our overall portfolio volatility is significantly lower than the markets. We never leverage the whole portfolio based on a single idea/position.
- The target feature we try to maximize, is "risk adjusted returns".
- Another goal is trying to "minimize trades". Earlier versions of the system were not viable/practical, because they required too many trades.
- It is short term oriented and doesn't try to predict what the markets will or won't do over longer time periods.
- It performs best when markets are volatile and uncertain.
- It is not a "fixed rule" system; it adapts over time.
- It uses many of the lessons we've learned over the years including many that I've written about in prior articles.
- By now, the system is largely automated, so we have to spend little time actually following it and executing the trades.
- The system is still evolving and we expect to improve it further as time goes by.
January was a very bad month for U.S. markets; the S&P500 dropped by 8.57%, the canonical ETF representing it (SPY) did a bit better: it dropped from 90.24 on the last day of 2008, to 82.83 on Jan 31, 2008 -- a 8.21% drop.
Our system has literally trounced the S&P500. It outperformed by about 18% (in just one month) and with lower volatility than the S&500 thanks to staying partially in cash for a large portions of January. It also suffered almost no draw-downs.
Since we were in an early experimental mode, we decided to follow the new system using only a small portion of our portfolios. This prevented us from benefiting from all the out-performance. Still, it is great to be firmly in the green in a month that has been so awful for the markets. In the coming months, I plan to gradually increase my exposure to the system as I gain more confidence in it. For February, I'm planning to double my exposure to the system compared to January.
Without further ado, let's look at some of the January stats as I recorded them.
# of round-trip trades (buy+sell or short+cover) 8 Gain/loss ratio: 7:1 (87.5%) Average gain (7 gainers): +10.88% Average loss (single loser): -2.39% Maximum gain/trade: 28.06% Minimum gain/trade: 5.48% Maximum holding period: 17 days (1 instance) Minimum Holding period: 1 day (2 instances) Median holding period: 3 days Average holding period: 5.875 days
The current algorithm approach, emphasis, and feature weightings, are designed to perform best at times of uncertainty and high volatility. The system, it is exact present form, is almost certain not to perform as well, once the bull market returns. As I wrote a month ago, I believe we are still far off from such possibility.
As always, this isn't intended as investment advice. It merely reflects my own thinking and actions at the time of writing. In the immortal words of John Maynard Keynes "When the facts change, I change my mind. What do you do, sir?"
Every investor should make up his own decisions based on his risk tolerance, comfort-zones, convictions, and understanding.
Any feedback is welcome.