Categories
Book Reviews

Rishi K. Narang – Inside the Black Box – Review

Rishi K. Narang - Inside the Black Box - ReviewA book about quantitative and high frequency trading. Published in 2013, this is the second edition. Rishi Narang talks about the quants, traders who use computerized approach to find alpha strategies to generate returns by systematically selecting and sizing various financial holdings. This book explains the difference between discretionary and systematic trading: what, how and why quants do what they do. While creativity comes from humans, computers are much better at repeatable action that needs strong discipline.

The quant universe

In this introductory chapter on why quant trading matters, the author writes about the quantitative world in general, how some hedge funds run these automated strategies and what their returns have been. High Frequency Trading (HFT) became more known in the public around 2010. This means huge amount of trades in a very short timeframe that a human being would not be able to execute without the help of computing power and trading algorithms. Rishi gives some statistical overview of how much of the markets’ volume at the time of writing (2013) was being traded by quants and their algorithms. The number is surprisingly high and I can only imagine it being even higher today in 2020.

What is a quant?

Rishi explains that while there might not be much difference in what a quant’s or a discretionary trader’s strategy does, the actual difference lies in how the strategy is implemented. Also, quant trading does not mean that human contribution is eliminated as the topic is about quants not robots. The computer can do as much as it is told to do. A black box is often referred to as something complex and secretive, but in reality it can be very transparent and logical. This book is about alpha-oriented strategies which means the returns are independent of any market direction in the long run. Strategies that follow some market index (like the S&P500) are beta-oriented.

Theory-driven vs data-driven alpha models

When someone is seeking alpha, there are only a small number of core strategies that can be implemented in numerous ways. Two types of science that quants use in their work are theoretical and empirical. Basically, theoretical scientists try to understand why something is happening and then test it. Empirical scientists believe that observing past data can help them find patterns to predict future movements without giving any explanation why such patterns exist. This is purely data-driven.

Most quants are theory-driven. Many often think that quants do something very complex and unique, but in reality the core concepts of trading systems overlap a lot. Usually, there’s six types of strategies: trend, reversion, growth, value/yield, quality and technical sentiment. The data being used can be either price-related or fundamental. These strategies in combination with data being used are further discussed in the book with examples.

Data-driven alphas are rare because they are more complicated. Data mining will possibly tell a trader if there are recognizable patterns to predict future movements that one could quantify and profit from. One of its advantages is that being more technically challenging it is being far less practiced. Another advantage is that through data mining one can find behaviors which have not been classified under any theory yet and the trader doesn’t need to understand the reason behind the behavior. This is why many HFT participants prefer the empirical approach.

The author goes through many important topics in his book about implementing quant strategies like forecast target, time horizon, bet structure, investment universe, model definition, conditioning variables, run frequency. These lead to a great variety of how many different quant strategies there can really exist. And a combination of different alphas is what gives diversification to a quant just like diversification is used in other aspects of financial life.

Risk models

Risk management doesn’t mean one should try to avoid risk or reduce losses. It is about the selection and sizing of market exposure to improve consistency and quality of returns. This part of the book describes how quant traders approach, measure and manage risks. Proper risk management should reduce day-to-day volatility of returns to the acceptable level for a trader but also help to reduce the probability of large losses.

Transaction costs

These are commissions and fees, slippage and market impact. Market impact is the change in price due to the trader going to the market with a large order. Seeing the difference between slippage and market impact can be tricky as they both depend on liquidity and affect the final execution price. However, these two may affect a quant much more than commissions and fees which are known ahead and can be better predicted into a black box. The author explains four types of transaction cost models: flat, linear, piecewise-linear and quadratic. A quant tries to apply a model that would best forecast the transaction costs but it’s usually hard to predict it right, so there needs to be a balance between accuracy and simplicity.

Portfolio construction models

These are rule-based and optimized models. The types of rule-based portfolio models are equal position weighting, equal risk weighting, alpha‐driven weighting and decision‐tree weighting. The inputs to portfolio optimization are expected return, expected volatility and expected correlation. There are many different types of optimization techniques and Rishi Narang has provided some of the more popular approaches in the book.

Execution

The purpose for good execution is basically to reduce costs of transacting. It’s about deciding the aggressiveness to execute based on the signal strength and confidence, trading goals. Aggressive execution is a market order that wants to get filled immediately at the best price available. Passive execution is a limit order that sets the best price to get filled at but with the chance of not getting filled at all. Rishi explains the times before electronic markets and how the algorithmic world operates today. He describes other order types used by quants.

Data

The author makes a point how important is the data that a quant uses for research. A quant needs to have quality input to the black box to get useful output. He goes on describing the timing of data, sources and types of data, cleaning and storing it.

Research

Here Rishi starts with a scientific research method and writes about idea generation with interesting examples. Next, he addresses many of the challenges of testing a trading system. The author gives helpful insights about in-sample and out-of-sample testing.

Beyond the theory

After two parts of quant trading theories, the book continues with a practical guide of quant strategies. Several real life examples how quants have struggled or blown up in specific market conditions makes me think about the pros and cons of quantitative trading compared to discretionary approach. Rishi provides a clear and good description on how to evaluate quants and their quantitative trading strategies. He has extensive experience interviewing hundreds of different quant traders and their methodologies. Reading the questions asked in these interviews, I tried to answer them myself about my own process and feel I didn’t do too bad, really.

The last, fourth part of the book is about high-frequency and high-speed trading, which at the first glance feel like the same but actually differ according to Rishi K. Narang. The author writes about controversy regarding HFT, why providing liquidity is actually good to everyone, how many of the accusations made by opponents of HFT are clearly false and provides some statistics supporting the fact that HFT has not changed the market volatility. He ends the book with ideas about the future of quant trading.

I got many good pointers from the book but I don’t want to get too much into detail, therefore I recommend to read this book if you are interested in quant trading. I am one step closer to optimizing my quantitative portfolio construction. Not that it’s something that ever gets “done” but still good to assume it’s getting better.

Share this post

Leave a Reply

Your email address will not be published. Required fields are marked *