Walking through the evolution of algorithmic trading

The concept of trading one possession for another has been around since the beginning of time. In its earliest form, trading was useful for exchanging a less desirable possession for a more desirable possession. Eventually, with the passage of time, trading has evolved into participants trying to find a way to buy and hold trading instruments (that is, products) at prices perceived as lower than fair value in the hopes of being able to sell them in the future at a price higher than the purchase price. This buy-low-and-sell-high principle serves as the basis for all profitable trading to date; of course, how to achieve this is where the complexity and competition lies.

Markets are driven by the fundamental economic forces of supply and demand. As demand increases without a commensurate increase in supply, or supply decreases without a decrease in demand, a commodity becomes scarce and increases in value (that is, its market price). Conversely, if demand drops without a decrease in supply, or supply increases without an increase in demand, a commodity becomes more easily available and less valuable (a lower market price). Therefore, the market price of a commodity should reflect the equilibrium price based on available supply (sellers) and available demand (buyers).

There are many drawbacks to the manual trading approach, as follows:

• Human traders are inherently slow at processing new market information, making them likely to miss information or to make errors in interpreting updated market data. This leads to bad trading decisions.
• Humans, in general, are also prone to distractions and biases that reduce profits and/or generate losses. For example, the fear of losing money and the joy of making money also causes us to deviate from the optimal systematic trading approach, which we understand in theory but fail to execute in practice. In addition, people are also naturally and non-uniformly biased against profitable trades versus losing trades; for instance, human traders are quick to increase the amount of risk after profitable trades and slow down to decrease the amount of risk after losing trades.
• Human traders learn by experiencing market conditions, for example, by being present and trading live markets. So, they cannot learn from and backtest over historical market data conditions – an important advantage of automated strategies, as we will see later.

With the advent of technology, trading has evolved from pit trading carried out by yelling and signaling buy and sell orders all the way to using sophisticated, efficient, and fast computer hardware and software to execute trades, often without much human intervention. Sophisticated algorithmic trading software systems have replaced human traders and engineers, and mathematicians who build, operate, and improve these systems, known as quants, have risen to power.

In particular, the key advantages of an automated, computer-driven systematic/algorithmic trading approach are as follows:

• Computers are extremely good at performing clearly defined and repetitive rule-based tasks. They can perform these tasks extremely quickly and can handle massive throughputs.
• Additionally, computers do not get distracted, tired, or make mistakes (unless there is a software bug, which, technically, counts as a software developer error).
• Algorithmic trading strategies also have no emotions as far as trading through losses or profits; therefore, they can stick to a systematic trading plan no matter what.

All of these advantages make systematic algorithmic trading the perfect candidate to set up low-latency, high-throughput, scalable, and robust trading businesses.

However, algorithmic trading is not always better than manual trading:

• Manual trading is better at dealing with significantly complex ideas and the complexities of real-world trading operations that are, sometimes, difficult to express as an automated software solution.
• Automated trading systems require significant investments in time and R&D costs, while manual trading strategies are often significantly faster to get to market.
• Algorithmic trading strategies are also prone to software development/operation bugs, which can have a significant impact on a trading business. Entire automated trading operations being wiped out in a matter of a few minutes is not unheard of.
• Often, automated quantitative trading systems are not good at dealing with extremely unlikely events termed as black swan events, such as the LTCM crash, the 2010 flash crash, the Knight Capital crash, and more.

In this section, we learned about the history of trading and when automated/algorithmic is better than manual trading. Now, let's proceed toward the next section, where we will learn about the actual subject of trading categorized into financial asset classes.