Forecasting and Timing Markets: A Quantitative Approach

[Note: This is newly published 2nd Edition updated in April 2021, full color version, which demonstrates that 8 out of 9 models beat buy-and-hold (bah) and average Joe's (aj) strategies significantly, based on backtest results with the past 15 years of historical prices.]

Stock market has been a mystery for over a century. From Dow’s theory to Elliott wave theory, many pioneers had attempted to formulate a theory that could help forecast and time market, but none had succeeded much. More recent legendaries such as Warren Buffett, Jim Simons, and so on, have made huge successes, but they could not be easily replicated. However, after years of research, I seem to have found a good alternative, i.e., building math models and letting machines do all hard work by running those models and generate signals based on well-tested historical patterns. Furthermore, I am finally able to turn a vague idea of my own I have had for years into a concrete math model named simple cascading indicator (sci), which could beat all classic models I tried most of the time. I felt so excited, so I recently summarized all my findings and new results into this 2nd edition of my FTM book. Out of all, a chart in Chapter 8 shows how my system has outperformed all three market indices by large margins, based on the historical data of past 15 years with 28 symbols. Many tables in the book also show that my models easily beat the buy-and-hold (bah) and average Joe’s strategies for each symbol most of the time!

This text attempts to fill the vacancy of how one can forecast and time markets more quantitatively. For this purpose, the author developed a model-based system, named AlphaCovaria, to help demonstrate how to use various simplest, readily available technical indicators to forecast and time markets approximately while eliminating subjective speculations at the same time. Centered on various math models, the author’s AlphaCovaria system has three main components: an AlphaCurve program for charting, a BTDriver program for running all backtests, and an AlphaCovaria driver for generating buy/sell signals based on symbol profiles learned through backtests. This kind of formula-driven approach is more promising for building more high-performance strategies.

The text is made concise and precise of about 100 pages only, as a working method does not need to be wordy. Math models, data and charts can help explain more effectively and convincingly. Finally, this 2nd edition of the book also shares my live trading experience using real money in my Fidelity and eTrade accounts with my AlphaCovaria system. Such data can be found nowhere else.

HENRY H. LIU, Ph.D., was once a physicist. Through his theoretical research and modeling on projecting trajectories of charged particles traveling at nearly the speed of light, he made unique, crucial contributions to several large-scale scientific research facilities in the national labs of China, France, Germany, and the United States. Later he jumped to computers and focused on software systems performance optimization, benchmarking and sizing. His highly acclaimed text, Software Performance and Scalability: A Quantitative Approach, has been in use for educating CS students worldwide. In more recent years, as the iPhone traffic forecaster as well as lead iOS systems performance and capacity planning engineer, he applied his past scientific predictive modeling experience as well as his extensive systems performance and capacity planning experience to helping forecast yearly iPhone launch events and get better prepared for dealing with high peak traffic. With this job, while delivering the highest degree of customer satisfaction with his fellow teams, he has consistently achieved high forecasting precision for iPhone launch unusually high-volume peak traffic, e.g., around 90% average forecasting accuracy for the latest iPhone 11 launch. His quantitative, methodical, profile and model-based approach has been proven to be widely applicable to forecasting and timing any events of time-series nature.