It is no secret that I like to back test every market observation that I have. I might not show all the data in these columns because most readers probably aren't interested in the details, but it would be rare for me to make a statement or a trade for that matter that wasn't grounded in the data. I have been doing it this way for over 10 years - observing, writing code, and testing.
This style of investing fits my personality well, but it was also developed out of necessity. I am self - trained, and I don't have the pedigree that other analysts might have. I didn't go to Wharton Business School, but I did go to Penn as an undergrad getting an English Literature degree. I have always felt that if I was going to take a position or have an opinion on this or that, I must come armed and ready. There is no better way to justify one's actions than having the data on your side. Opinions grounded in the data create conviction. I also found over time that my quantitative trading models work a lot better than my trading decisions that I make by discretion. There was a reason why I back tested: it worked.
So what got me to thinking about this the past week? It was my position in the i - Shares Lehman 20 plus Year Treasury Fund (symbol: TLT). In the last 30 days alone, I have been "very constructive", "still constructive" and "less constructive" on the TLT. When it comes to bonds, I use a fundamental and technical model. The fundamental model is based upon over 40 years of data and relies upon inter-market inputs such as commodities. The technical model generates the buy and sell signals provided they are in alignment with the fundamental model.
My less constructive stance on TLT last week was based upon the technicals. As explained, it appeared that the breakout on the daily charts had turned into a fake out. Mind you, my parameters for failure, which I discussed previously, had not yet occurred, but it was looking like it was going to happen. Last week I did mention something about a low risk entry point as prices had backed up to support but the statement seem to lack conviction. I stated:
"The good news essentially is that TLT has a low risk entry point. See figure 2 a weekly chart of TLT. In other words, prices are sitting against the 89.17 support level. Of course, a good low risk entry point doesn't say anything about the potential for a price gain...."
Based upon the "potential" break down in price and the fact that the fundamental model was looking weak throughout the week, I sold one half my position in the TLT. Of course, this was before any signals were generated in the models and of course, it was also at the low price point for the week. Not good!
So how did this happen? As I generally use weekly data, I was trying to anticipate (i.e., use discretion) a potential technical and fundamental change in my models. Second, reports that uber- investors Warren Buffet and Bill Gross were bearish on bonds had me thinking that may be these guys knew something that I didn't.
So what is the moral of the story? The models still remain better than my own discretionary abilities -- this is why I invest using models!! Opinions should be ignored. Bill Gross and Warren Buffet are worth listening too, but they should not figure in my trading decisions especially since I didn't build their inputs into the models.
The good news? I still have a half position in TLT. So I was only half bad.
For the record, we have closed the week with the bond model bullish - slightly less bullish than the previous week but bullish nonetheless. The technical picture has obviously improved; support or failure levels remain at 89.17 on the TLT.