After some relaxing reading, time to get serious again. As such, I choose one of the topics that remain as a myth to most traders. Well, hopefully this book can provide more insightful stuff to me...
Well, despite the facts that this book was published many years ago; it still serves as a bible or at least a starting point for those who seek to understand quantitative trading more. Having said that, this book is not the know-how or technical stuff about algorithm. This book is conceptual rather than technical. However, for those with basic quantitative trading knowledge, this book is good enough in thought-provoking and I benefited a lot via this book.
The quotes below good enough to support my point:
If a quant is good at forecasting volatility or dispersion, there are far more interesting and productive ways to utilize these forecasts (for example, in the option markets) than there are in a risk model that governs leverage. A more theoretically approach seeks to increase leverage when the strategy had better odds of winning and to decrease risk when the strategy has worse odds. The trick, of course is to know when the odds are on one's side.
Statistical risk models are subject to being "fooled" by the data into finding risk factor that will not persist for any useful amount of time into the future. It is also possible for a statistical risk model to find spurious exposures, which are just coincidences and not indicative of any real risk in the marketplace. This is a delicate problem for the researcher.
Risk management is frequently misunderstood to be an exercise designed to reduce risk. It is really about the selection and sizing of exposures to maximize returns for a given level of risk. After all, reducing risk almost always comes at the cost of reducing return. So, risk management activities must focus on eliminating or reducing exposures to unnecessary risks but also on taking risks that are expected to offer attractive payoffs.
Too much emphasis on the opportunity can lead to ruin by ignoring risk. Too much emphasis on risk can lead to under-performance by ignoring the opportunity. Too much emphasis on transactions costs can leas to paralysis because this will tend to cause the trader to hold positions indefinitely instead of taking the cost of refreshing the portfolio.
The instability of correlations among financial instruments is more or less a fact of the world. It is not the fault of optimizers, nor of correlation as a statistic, that this happens to be the case in the financial industry.
Near-neighboring sets of parameters should result in fairly similar results, and if they don't researcher should be a bit suspicious about them, because such results may indicate overfitting.
Models that are parsimonious utilize as few assumptions and as much simplicity as possible in attempting to explain the future. As such, models with large numbers of parameters or trading signals are generally to be viewed with skepticism, especially given the risks of overfitting.
We find that the statement that quants underestimates risk is likely to be true, but we also find this to be due more to human nature and the circumstances are rare events than to something specific in quant trading.
Longer horizon quant strategies tend to go though longer and streakier performance cycles. They can outperform or under-perform for several quarters on end, and it can take several years to evaluate whether there is really a problem with the manager. Some longer term strategies also demonstrated conclusively that they are subject to crowding risk. Short-term strategies, by contrast, tend to be very consistent performers, but they cannot handle much capital. They are therefore very desirable but also not always practical. Furthermore, when one find a good short-term trader, it is not clear that the trader will remain small. Many traders are tempted to grow their assets.
For a rating of 10, I am going to rate this book at 9. Despite the excellent part, this book is a bit out dated with current algorithm environment. Algorithm evolved a lot these days and the evolve part would not stop in the future. As such, I am very much looking forward for the new version of this book. (Better still if the author decided to write a new book about quantitative trading) Overall, this is an excellent book and I highly recommend traders to explore it.