« Altera apologizes, Thursday, July 28, 2005, 5:02 PM | Main | U.S. bonds are weak, Friday, July 29, 2005, 10:56 AM »
July 29, 2005
A proven mechanical trading system, Fri., July 29, 2005, 8:57 AM
One of the positive aspects of blogging like I do is that readers often share their ideas. I recently received the following mail from Australia (via Spain), from a software expert (Keith) who is developing a personal-use auto trading system. He has been using the non-automated, mechanical (i.e., "rules"-based system), described below, for years, and now uses it as a benchmark in developing automated systems.
G'Day Bill,
I thought I'd drop you an email about a simple trading system I have been using for years.
The strategy:
Once a month:
1) Consider only the universe of stocks ranked "Timeliness 1" by Value Line. Why use Value Line "Timeliness 1"?
2) Sort universe by Relative Strength 26-week period to get top 20 stocks ranked.
3) Buy top 10 stocks in equal amounts
4) Sell a stock when it drops off the top 20 list and replace with the next one in the top ten not already owned. (optionally rebalancing funds to weight equally).
-------------------
This method has been backtested to 1969 with returns that approximate a 30% compound annual growth rate (CAGR).
The above "screen" was developed as a community effort by a group of investors calling themselves "mechanical investors". To compare apples to apples between the myriad of trading strategies, the following measures are generally used:
- Geometric Standard Deviation (GSD) of annual returns. Measure of volatility-draw down. Notes and example program.
- Compound Annual Growth Rate.
- Sharpe Ratio.
The above simple strategy "RS26wk Hold Till Drop 20" has a CAGR of 30%, GSD 34 and Sharpe Ratio around 0.8. My live returns have approximated these figures.
-------------------
Interestingly and as a kind of confirmation, Ford research also recommends a similar method with similar results, in their report to fund managers: "Alternative Portfolio Applications of Value Momentum Model". I have put it here on my website for your viewing here.
There are many tweaks and alternative screens aimed to maximize CAGR and minimise the GSD. However the more complex they get, the more danger there is that the backtest results are just curve-fitting the historical data, so simple and robust screens generally perform in real time closest to the backtested returns. As expected, higher CAGR means higher GSD, and past GSD is a good predictor of future GSD.
Not all the methods employed by mechanical investors fit my personal style: They generally shun most other technical analysis methods, especially any that cannot be reproduced mechanically and thus backtested on more than 20 years of data. They only check their portfolios once a month and for the time it takes to blindly buy/sell the stocks appearing in the screens. I simply use this as the yardstick to beat the market.... to do better than 30% annualized return on 30 minutes a month work, proportional to the effort and cost it takes to build such a auto trading system.
/Keith
I have a lot of views on automated trading systems (aka algorithmic trading) because I have built them in the past, and am presently building another (which is taking longer than anticipated).
The bottom line is that most of them work, but not as well as first planned, and they are a lot like an ultra-sophisticated Formula One racecar that still needs a competent driver if the goal is to win the race.
Besides, the racetrack frequently changes, requiring a different set-up, and different driving tactics, to be able to continue to win.
If success with algorithmic trading were assured, then every major firm on both the buy-side and the sell-side would be beating the market " because they all use them today. Shocking, perhaps, but probably 60+ pct of all trading (volume or value) on the NYSE is generated by these systems.
And, since we now live in an age of robots, where independent thinking is discouraged, you can bet that use of algorithmic trading systems will grow.
On another point, the author of this e-mail (Keith) has noted the value of Value Line research services. All my readers know how valuable a tool I believe Value Line is. I give it a 5-star rating, worthy of going to the public library to access the full service (over and above the free Dow 30 reports), if you cannot afford the cost.
Still, Value Line ought to be just one tool in your toolkit. Over the years, you need to be constantly adding tools.
I want to thank Keith for contributing his knowledge to this blog. Like most of my readers, I think you can see he is serious about trading securities as the key to building wealth.
Posted by Posted by Bill Cara on July 29, 2005 08:57:40 AM | Category: Learning Center , Trader Tools , Value Line Research
Bill,
I continue to love your blog. Here's a dumb question regarding the 26week rsi mentioned above. Would a 26week rsi be the same as a 130 period rsi on my chart service?
Posted by: blslay at August 1, 2005 4:06 PM [link]