OK, I'll admit it. None of us are investment gurus and we haven't made (or lost) millions playing the stock market. However, we've been asked a few times to work on projects related to automated trading systems and these are some of the books we found most interesting or useful in understanding the problems involved. Click on the titles to see reviews by other people and other information from your local Amazon.
If you can survive reading the title, this book is actually an anecdotal history of the Kelly Criterion — a formula which identifies how much should be wagered on a risky undertaking. A surprising cast of characters is involved, from underworld figures to physicists and mathematicians, and from 1738 (not a misprint) to the present day.
This isn't a mathematical enough treatment for me, but it's well worth reading to find out how people have tried and succeeded (or failed) in exploiting Kelly's (and Bernoulli's) work.
(If you need to play around with numbers, the Kelly Calculator may be useful.)
If you've ever dreamed of creating a computer system to beat the bookies or the stock market (and who hasn't?) then Steven Skiena's book is for you. Skiena describes his own (and his team's) efforts to create an automated system which would place winning bets on jai alai, a Basque game which is also played in parts of France and some cities in North America. The book describes the game itself (a game with similarities to tennis, squash and rugby fives), the convoluted methods by which tournaments are held (a sort of round robin) and the pari-mutuel gambling system which is used to place bets on the outcome.
To create the perfect system, Skiena and his team needed to model the tournament structure, the effects of player skill on match outcomes,and (since this the odds are offered using a pari-mutuel system) the betting habits of the general public. They then needed to identify bets which would (on average) be profitable.
The team succeeded, the program they developed ultimately succeeding in returning about 500% on its initial gambling stake in a single year. (The bad news, as Skiena points out, is that it would not be possible to use such a system to bet large amounts of money on jai alai, since such bets would significantly depress the odds available).
A must read for anyone seriously interested in "beating the system".
Lacking the modesty or limited objectives of either Taleb or Paulos, this book claims it will reveal ... how you can use your computer to gather, analyze, and detect profitable market inefficiencies — the key to making winning trades day after day.
Despite this publishing hyperbole, there is much to recommend this book if you are planning on implementing a trading system. There are explanations of neural net, technical analysis, and data mining approaches. There are also descriptions of two or three major projects, discussions of possible commercial data sources, and key questions to ask in evaluating systems.
Some of the chapters in this book are somewhat dated: with the substantial changes in hardware and software since 1999 we thankfully don't have to worry anymore about whether incoming data will overflow our 16550 UART. However, many design considerations never change and this book is still worth a read if you are considering designing or evaluating an automated trading strategy.
Paulos lost money on the WORLDCOM crash, and uses this as a jumping off point for explaining the mathematics behind the stock market. There are good qualitative explanations of many of the numbers and theories used in attempting to predict the price of stocks. I especially liked his description of the Efficient Market Hypothesis Paradox (if everybody believed in it, the Efficient Market Hypoethesis would no longer be true) and its imaginary opposite, the Sluggish Market Hypothesis.
Another good point he makes is that the criteria for success of a trading system is not whether it makes money. That's only a necessary condition. A successful trading system must make more money than simply investing in treasury bills, or buying an index fund.
The book covers in outline such subjects such as technical analysis, beta, portfolio theory, etc.
Although there are some excellent nuggets of information here, I found the writing style distracting. I was also disappointed that another book by the same author (Once upon a number : the hidden mathematical logic of stories) covered much of the same ground with the same examples.
This book contains musings on random events and its effects on the market (and life in general) by a professional trader, Nassim Taleb. There are thoughts here which I found quite profound concerning the nature of inductive logic (reasoning from events to rules), as well as interesting examples and explanations of how we allow ourselves to be fooled by random phenomena.
Taleb is particlarly fascinated by what he describes as the Black Swan Problem. We see lots of swans. All of them are white. We infer that all swans are white. Unfortunately we have never been to Australia, where the swans are black as well. If we build our trading systems on such principles will the appearance of a black swan wipe us out?
The style of writing here is collection of literate musings and digressions which I rather liked but, judging by Amazon reviews, it appears to irk some readers.
One of the key assumptions made by many investment and pricing theories (e.g. Black-Scholes) is that prices behave in a manner which can be modelled by the normal distribution. It seems on the face of it a reasonable assumption, but it isn't.
Mandelbrot makes a strong case that the normal distribution is a poor estimator for the distribution of price changes, and that a fractal distribution paints a more accurate picture. Why does it matter? If you take the normal assumption then large jumps in price are extremely unlikely, and price changes on any day are not affected by changes the previous day. In fact the extremes do happen, and the magnitude of today's changes does correlate with previous day's trading. The net effect is that the normal distribution assumption seriously underestimates risk.
This book is in the popular science category. There is no deep mathematics or equations. (A shame, if you are mathematically inclined — good news, if you aren't.) There's a reasonable mix of anecdote and experimental evidence, but a little too much schadenfreude when Mandelbrot manages to prove himself right and others wrong.
This book won't make you rich, or give you a sure-fire way of making money on the markets — it's author carefully points this out. What it will give you is a deeper understanding of the shaky foundations on which portfolio theory, option pricing and many other models are based, as well as some ideas on how to more realistically test a trading system on simulated data.
Judging by this auto-biography Edwin Lefèvre (a pen-name of Jesse Livermore) may have been the sort of person your mother warned you against.
He made his living from speculating on the stock market, sometimes using techniques which would probably be illegal today. He both won and lost money, and this book contains his insights into successful and unsuccessful trading periods. Here's one of Lefèvre's asides on the nature of speculation, which follows a story of a man who had a successful system, thought he could do better by changing the system, then wiped himself out:
I sometimes think that speculation must be an unnatural sort of business, because I find that the average speculator has arrayed against him his own nature. The weaknesses that all men are prone to are fatal to success in speculation — ususally those very weaknesses that make him likable to his fellows or that he himself particularly guards against in other ventures of his where they are not nearly so dangerous as when he is trading in stocks or commodities.Perhaps an automated system will not be subject to such human mistakes? Or will it, too, lack the humility of traders such as Taleb?