To: gonzongo who wrote (8959 ) 3/13/1999 6:59:00 PM From: Nine_USA Respond to of 11149
Andy, <<I would like to caution you- I often find that by accidentally weighting a db with "good" factors- such as current trend etc.- my back testing gave erroneous data- It was only when I clearly eliminated all forward information and relied on historical information only ( ex: avg volume then vs current avg volume) - that I was able to get accurate results.>> I couldn't agree with you more! That is why for each market day, I create a mini datafile for all stocks in QP2 to reflect the data available to the investor looking to invest the next day. When I backtest, for a 1 year holding period starting Mar 11, 1998 to today (for example), I do all ranking strictly on the contents of this minifile I created back on Mar 11, 1998. This avoids problems for split affected prices, volumes, moving averages, total shares out, # short, and hopefully, all the rest. Essentially, I am working with a miniature of what QP2 for that date, contained that I use in my approach. For obtaining the forward investment returns, I reference the minifile I created for March 12, 1999. I use this file to get the prior 12 months % price changes for all stocks. There is a small problem in that some of the year ago stocks are no longer present due to name change, merger, bankruptcy, exchange change etc. But this is only a sprinkling relative to the total database. I started out looking at optionally limiting my starting population of stocks to the OEX 100, or the SP500, or the Russell, or industry groups, etc, but discarded this as impractical because these groups change members from time to time, so I would need to keep each list group for each market day. I now do 'Value' weightings using all the 4000+ QP2 stocks which possess the requisite data. <<You could test a period when the RUT was moving more strongly- using a smaller db from earlier years- and see what happens with the results.>> I can't do this because my approach was fully formed on Feb. 17, 1998 and I have these mini datafiles only for each date since. I will have to depend on future data to further validate and refine. I have been putting in 50 to 60 hours per week for the last 2 years pursuing what I want which is a mechanical, non-subjective method for investing. This is something which I feel may have some entrepeneurial value and for that reason I am not prepared to discuss details of my approach. I can discuss what I have found in general as to some variables, market sectors, and holding periods.