Artificial Intelligence in Finance

Part IV. Algorithmic Trading

Success means making profits and avoiding losses.

Martin Zweig

Part III is concerned with the discovery of statistical inefficiencies in financial markets by the use of deep learning and reinforcement learning techniques. This part, by contrast, is concerned with identifying and exploiting economic inefficiencies for which statistical inefficiencies are a prerequisite in general. The tool of choice for exploiting economic inefficiencies is algorithmic trading, that is, the automated execution of trading strategies based on predictions generated by a trading bot.

Table IV-1 compares in a simplified manner the problem of training and deploying a trading bot with the one of building and deploying a self-driving car.

Table IV-1. Self-driving cars compared to trading bots

Step	Self-Driving Car	Trading Bot
Training	Training AI in virtual and recorded environments	Training AI with simulated and real historical data
Risk management	Adding rules to avoid collisions, crashes, and so on	Adding rules to avoid large losses, to take profits early, and so on
Deployment	Combining AI with car hardware, deploying the car on the street, and monitoring	Combining AI with trading platform, deploying the trading bot for real trading, and monitoring

This part consists of three chapters that are structured along the three steps, as illustrated in Table IV-1, to exploit economic inefficiencies through a trading bot—starting with the vectorized backtesting of ...