High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems (Wiley Trading) [Hardcover]

This book tries to show the current state of algorithmic trading -- although it confuses algorithmic (i.e. programmatic), high-frequency, and microstructure-driven trading. Sadly, the shortcomings in the text will be apparent to anyone with experience in the field. Very little empirical work backs up the assertions; and, no theory is incorporated about strategies and tactics. Even the rare theoretical notes suffer from inconsistent notation: hats and epsilons are added or omitted as though they are mere decorations. Further, there is much to disagree with. When a book mentions the current top high-frequency traders and omits Getco or Lime Brokerage/Tower Capital, you know something is wrong. Another example: The book repeatedly confuses the speed of execution with the investment holding period.

I had hoped to use this book for a market microstructure and trading course I teach. However, this book does not approach the standards: Larry Harris's Trading and Exchanges; Joel Hasbrouck's Empirical Market Microstructure; and, Maureen O'Hara's Market Microstructure Theory. O'Hara's book serves as a proper contrast: While the book is nearly 20 years old, the basic tone and approach are still sound. This book, however, focuses too much on the "here and now" and misses the larger issues behind the move to electronic and algorithmic trading. Further, these issues are not new; many are mentioned offhandedly in Edwin Lefevre's classic Reminiscences of a Stock Operator (written in 1923).

Aside from the flaws, I am sure some will find this book informative. However, I doubt any serious quants or academics will find it agreeable. Overall, the perspective is too narrow yet is proclaimed to be broad; the ties to theory are weak to non-existent; some minor parts are wrong; and, I found little that was both good and new.

I'm not sure who or what this book is written for. It obviously requires some mathematical and financial background. It also very obviously isn't useful to practitioners; not even larval ones -this despite it being alleged as "a practical guide." Could it be a textbook for academics? Why would an academic need a definition of returns ... and in chapter 8? I'm guessing it's more of a review for educated types who want to know more or less what HFT is -it succeeds best at this level, though the profusion of equations and literature citations doesn't make it real approachable at this level either.

The book is a sort of review article; a good fraction of the text consists of references to other texts and articles. There are 20 pages of references for a generously spaced 300 page book, and probably another 20 pages of text dedicated to citing these references. I don't think it's a very good review of the subject in the title. While the chapter outline is promising, and something like what a book on HFT should contain, the meat inside the chapter headings is thin and gristly, with few vitamins and minerals. Mostly it ends up reading like "Bob (2006) says blah." Much of what is referenced is irrelevant, much of what is relevant is never mentioned, and some of what is stated in the text is laughably wrong. I don't know who proof read this thing before it went to the publisher; whoever it is is either a moron, or wishes the author ill. I'm no HFT expert either, but I could have written a better book than this.

In detail: The first four chapters are pretty much useless to anybody I can think of, other than reporters and pointy headed panelist types who want to sound clever in cocktail conversations. Chapter five goes over some a standard bestiary of performance metrics. Chapter six is a very limited description of order types -again; something useful to a stuffed shirt over cocktail wieners: utterly useless, almost laughable to a practitioner. Chapters seven and eight has some weak stuff on "how to find signal" in financial noise. It's useful if you never heard of autocorrelation before, but it's not going to help you if you're in that sad state. A book on signal processing or econometrics might help you. There is also evidence that the author has not actually used some of the models and techniques she name drops here: for example, I've never heard of neural nets having the advantage of "significantly speeding up execution of the forecasting algorithm." Every neural net I've ever encountered sucked at speed compared to, say, regression or ARIMA. The section on tick data is weak. Tick data is certainly mentioned, and you'll know what it is at the end of the chapter, but the chapter doesn't actually tell you anything about "working with tick data." There's lots of tricks to ticks; none are covered here. Chapter 10 on market microstructure contains a section on the gamblers ruin which appears to be completely wrong. Either that, it's misprinted, or the rum I needed to get through this review is causing me to see things. It's probably not worth mentioning the equations are mislabeled in the text, but even in my liquored state, I noticed. Chapter 11 is sort of OK, though, for example, Joel Hasbrouck's chapter 11 on the same subject is much more informative. I also would have liked to have seen more and different stuff in there. I mean, what is the utility of including quotes from the Economist on Bayes theorem? I'd rather quotes from someone who knows what they're talking about, rather than quotes from some ding dong journalist, or perhaps some more information about market microstructure models. Chapter 12 on event arbitrage is good in that exists, and gives a rough outline of how it works. Event arb is awesome, because even schmucks like me can do it; I love event arb; nobody talks about it, but the author of this book actually did. Is this chapter useful to a practitioner? Nope. Again; good for the cocktail crowd and newspaper reporters: useless to anyone who wants to trade. Chapter 13 is on everyone's favorite fancy pants trading strategy, "stab art" (and, BTW physics is a hard science, not a "hard" science like it says in this book). "Statistical arbitrage" is a phrase which tends to cover a lot of different stuff; some of which is never discussed here: pairs trades, for example, is the standard textbook example; MIA. Triangle arb is also conflated with stat arb, which is silly and wrong: there is nothing statistical about triangle arb; triangle arb is just plain arbitrage. I guess there is some general hand wavy material which might give a hint to people looking for opportunities, but there is so much missing, it probably reads like greek to the uninitiated. Chapter 14 purports to be about portfolio management for HFT, though really, it's about classical portfolio management, which isn't so useful for HFT. HFT portfolio management is a genuine black art; a real chapter on it revealing some practitioner secrets (or even some decent academic references) would have been invaluable. This chapter also contains such bloopers as confounding Bayesian self-correction with genetic algorithms on page 209, which is sort of like a chef confusing an eggplant with a blender. It does contain an outline of the portfolio optimization technique used by the author, which I guess is vaguely sensible, but is rather ad-hoc and not particularly convincing, and has little to do with issues which arise with HFT portfolios. I take umbrage with chapter 15 on back testing. Back testing is hugely important in any kind of forecasting algorithm; waving your hands over the MAPE formula is pretty much useless. Nowhere is the sin of data mining given the attention it deserves: things like the bootstrap or establishing p-values for overfitting probabilities ... well, maybe I am expecting too much from a book which confuses eggplant with blender. Chapter 16 is supposed to be on implementation. FIX gets mentioned at least. So is C++, and, erm, Java. I do know of one firm which uses Java, but I know a lot more which use Matlab or Python (yes, even in HFT). A few concepts in software development and QA are mentioned in passing. I'm not sure who she's mentioning this for: any nerd who has slung some C knows what unit testing is. I'd rather a few paragraphs on issues with using time series databases. Chapter 17 (the liquor is getting me pretty hazy at this point) is on Risk management. She describes some stuff on using Pareto-Levy distributions to characterize tail risk. I know Taleb and company seem to recommend this, but I've honestly never heard of anyone trying to do it, because fitting Pareto-Levy distributions sucks. Probably some people do, but "I never hoyd of da bums." She describes some bootstrappy way of doing this, which I'm guessing is somewhat more useless than using the bootstrap to see if your dumb trading strategy has any legs to begin with. Trying to pin the tails on the Pareto distribution seems like a bad idea to me: I think of tail risk as consisting of things which you can't really model. The bits on "legal risk" and "operational risk" are good in that these words exist in the book, but bad in that they contain no actionable information: "talk to a lawyer" isn't really helpful to me. They also don't belong right before a section on stop loss orders, which is a really different kind of risk. The remaining chapters have turned into an alcoholic blur for me, but ch 18 "executing and monitoring" is worthless unless you didn't realize trading systems have to be executed and monitored, and at least chapter 19 finally mentions implementation shortfall and VWAP, which are kind of the harmonic oscillators of algorithmic trading at least. A lot of people would put these somewhere at the beginning of the book.

I don't know if anyone will read that far: I kind of wish I hadn't. The hangover is going to be epic and painful.

I was looking forward to this book but it's a big disappointment. I've worked on proprietary trading desks for over 10 years and have done all aspects of high-frequency trading. This book is not written by an expert - she just lists stuff that might be useful - and there's little hint she's done any of the work she talks about. If you're in the industry save your money. If you're aspiring to join the industry you might be worse off after reading it. The book seems more like a commercial gimmick for the author's career and current employer than a genuine effort to add value to the subject.