Slauson's Slant on Trading
Forecasting with Grandpa's Trick Knee
Contributed by John Slauson
Since the late 19th century, weather forecasters have measured barometric pressure to predict the weather. The earliest barometers were invented in the 17th century. They were made of a basin of water and a glass tube. As the atmospheric pressure changed, the level of water in the tube fluctuated up and down. Since then, more accurate methods of measuring barometric pressure have emerged. Even so, my grandpa swore that changes in the pain level of his trick knee were more accurate than any of the fancy instruments used by meteorologists.
As technology has advanced, so has the reliability of weather forecasts. Forecasters have developed a wide assortment of new tools. They feed data into complex computer models that provide increasingly accurate forecasts. The key to their accuracy is using a wide variety of tools to analyze the data such as barometers, radar, satellites, and weather balloons. If their computer models were limited to data from a single tool like a weather balloon, then almost certainly their forecasts would be pretty dismal.
Forecasting the weather and forecasting the markets have many similarities. I learned an important principle from John Bollinger many years ago that I later used in the ICE add-on for MetaStock. Mr. Bollinger emphasizes the importance of avoiding technical indicators that have collinear variables. What does this mean? Here's how he explains it: "A cardinal rule for the successful use of technical analysis requires avoiding multi-collinearity amid indicators. Multi-collinearity is simply the multiple counting of the same information. The use of four different indicators all derived from the same series of closing prices to confirm each other is a perfect example."
In practical terms for MetaStock users wanting to build reliable trading systems, it means they should use indicators that measure a variety of market behaviors rather than many that measure the same behavior. For example, the two most popular technical indicators are RSI and Stochastics. With only slight variations, these two indicators are almost identical. The chart below shows a 14-day Stochastic and a 14-day RSI along with the correlation of the two in the top inner window. Note: the correlation coefficient is extremely high, ranging between 0.80 and 0.90. They are almost perfectly correlated. Even without using correlation, it is visually obvious that the two rise and fall in unison.
While two indicators being highly correlated (like RSI and Stochastics) does not necessarily mean they cannot be a valuable part of a trading system, in this case with the mathematical formulation underlying each indicator being so similar, you can be sure that there is little to be gained using both in the same system. So just like weather reporters, you should use a variety of different tools, not variations of the same tool.
One of the reasons we get sucked in to using collinear indicators is they optimize very well. The optimized results of a set of collinear indicators almost always show better historical performance than does a set of non-collinear indicators. Why? Because it is much easier to "curve fit" a system comprised of three highly correlated momentum indicators than a system comprised of different categories of indicators. A system that is curve fit to the past will rarely perform well in the future. Unfortunately you can't make money trading the past.
A better approach is to combine indicators that measure different dynamics such as momentum, trend, volume, and volatility. When building a trading system consider using indicators from at least three of these four categories. The table below shows some of the MetaStock indicators categorized into these four categories:
For example, RSI, Chaikin Money Flow, and Bollinger Bands when carefully combined into a system will provide non-collinear input that will likely perform better in the future. RSI measures momentum, Chaikin Money Flow measures strength using volume, and Bollinger Bands measure volatility.
And if you can figure out a way to work grandpa's trick knee into your system, then perhaps you will have found the holy grail!
About John Slauson
John Slauson began his career with MetaStock in 1988. In 2000, he left and started Adaptick, a company that provided training and developed popular MetaStock add-ons ICE, FIRE and PowerStrike. Over the years, he's worked closely with industry experts like John Bollinger, Steve Nison, John Murphy, and Greg Morris. In 2008 he returned to MetaStock as a Product Manager.
No comments:
Post a Comment