Day Trade With AI: Theoretical foundations for discretionary, algorithmic, and diversified trading with AI

For many, a book on day trading is akin to a book on gambling. In the field of gambling theory, there are two notable books: “The theory of gambling and statistical logic” by Richard A. Epstein, published in 1977, and “The mathematics of gambling” by Edward O. Thorp, published in 1984. With the rapid evolution of AI, I’ve become curious about whether the gambling nature of day trading to profit from an efficient market has changed. To the best of my knowledge, in the field of algorithmic trading theory, I’m not aware of any notable books. James Simmons has been very successful in this field, but his strategies are highly proprietary. Tang’s book, when it appeared, satisfied my curiosity in this aspect. It is truly audacious and innovative that Tang has brought such a book on theoretical foundations of day trading with AI to the public.Tang created the theory of day trading with AI by pushing beyond the boundaries of Harry Markowitz’s Portfolio Theory, Eugene Fama’s Efficient Market Hypothesis, Daniel Kahneman’s Prospect Theory, Richard Thaler’s Behavioral Economics, and Andrew Lo’s Adaptive Market Hypothesis, in combination with Yann LeCun’s deep learning frameworks. For instance, it is known that Markowtiz’s Portfolio Theory only applies in a theoretical space when the stocks are uncorrelated and have constant variances. In the real world, however, such assumptions are not necessarily true. Therefore, we must hold a well-diversified portfolio for the long run to combat market volatility. It is brilliant that Tang extrapolated Markowitz’s Portfolio Theory to the realm of day trading. If the prediction window is narrowed down to as short as minutes, there is no way that stocks’ price actions are correlated, thus unleashing the power of diversification.As a believer that historical stock prices do not contain predictive information for AI to learn, Tang did not hype up any pattern strategy. Instead, he scrutinized every pattern strategy for psychological soundness and placed an emphasis on diversified trading with AI. This includes stock diversification, model diversification, and time diversification to secure an edge toward success. This approach leaves a lot of room for flexibility in training because a trading suggestion is not made by one single strategy, which could be subject to overfitting to historical data, but by a “committee” of experts. The insights delivered in this book guide not only day trading in the stock market but also many other machine learning applications.The book is written in an academic textbook style, complete with sections and end-of-chapter exercises. It is a perfect read for beginning and intermediate-level day traders with some programming knowledge and a curiosity about employing AI in their ventures. There is also a dedicated website for this book where readers can ask the author questions directly and download book resources for free. I asked Tang a question on a code example in the book via the book website and he responded promptly to get my problem solved.As for the printing quality, the book is printed in full color on premium paper. I appreciate that the color scheme, including all words in bold and section titles, is in green. After all, we all want a green day, don’t we?However, two shortcomings of the book are worth noting: (1) A software system based on the diversified trading theory is essential to fully satisfy readers’ curiosity. However, this content has been moved to the next book in the series. I understand this is a marketing strategy, and I will buy that. (2) The code for downloading financial data is based on the use of the Finnhub API. However, Finnhub has recently changed its pricing policy and it is no longer free. While readers can easily find free API services from other financial data providers, this part needs to be updated in the next book in the series.

I liked this book a lot. Before I make any comments, I must add that I have certain biases towards trading. Namely, I have always looked at it as a form of gambling. The individual investor is at a distinct disadvantage in that she or he doesnt necessarily have timely access to all of the data, the computational wherewithall, nor the deep pockets of an institutional investor. That being said, the author assembles a very credible approach to using machine learning and day trading. During the course of the book, he brings together portfolio theory, behavioral economics, adaptive market hypothesis, stock diversification, machine learning model diversification, and time diversification in order to secure an edge toward trading success. The book is written like a textbook with end of chapter exercises. The code provided is well documented and easy to follow. The book is clearly written and not dumbed down. The writing style is enjoyable to read and the authors enthusiasm for the subject shines throughout the book. There is a web site for the book and its follow on book. I asked the author a question about some of the code and he responded promptly with helpful information.