I meant to short S&P (the one that was out performed as a benchmark by 35% even in the bear market) , and not the bottom rank .
Well this can definitely be done. This may be one method of making the model market neutral. However, should it be done through simple dollar amounts or by incorporating the picks volatility as well (for example, say the model shows a beta of 1.5 so should we short a proportionally more of the benchmark compared to the long the picks)?
I have a question . How can you backtest the entire universe of stocks using technicals AND fundamentals? Is it even possible to backtest a technical model on all US stocks ? Which program will do that ? OR do you pick the stocks that had the best fundamentals then backtest your technical system on each of those?
None of the above. There is a simple criteria to determine the model's universe (approx. 2000 stocks at the present). The criteria is simple volume, capitalization etc. After that the model generates a list based on fundamental and simple momentum criteria (no advanced technical analysis....simple high/low over 52Wk etc). To backtest I had to do some coding on VB (nothing incredibly hard...I am a novice programmer, I did recieve help on some of the code from a friend of mine).
Well here are some monthly backtest graphs for equal dollar Long model, short benchmark. The equity curve (Jan 1989-Dec 2005) And here is the month return chart (Jan 1989-Dec 2005)
granville>I got the data from a friend as a database. I programmed the backtester in vb (pretty simple program but did get help on some of the tricky coding). It spits out the output in excel. Here is the long model only performance: Equity Graph (Jan 1989-Dec 2005) Monthly returns chart (Jan 1989-Dec 2005)
nice... notice that from time of "back to normality" (fundamentals does matter , dividend does matter...) starting from 2001 , you don't see big downs months and your queue clearly stars to outperform the market.
Yes. The long model short benchmark, seems to have lower monthly return volatility than the simple long model alone. But it doesnt have the blockbuster months like the long only system.
I did some reading along these lines and found a book "What Works on Wall Street" to be very informative. The author looks at base rates and uses stats to develop multiple systems with holding periods of 1 year. Maybe the book could be of some benefit to you: http://www.amazon.com/gp/product/00...102-1131692-8764927?n=507846&s=books&v=glance