The simple answer is “machine learning,” which consists of three general steps:
- Quants input algorithms or rules relating to market trends into a learning bot.
- They feed the bot raw data about the market, which it then uses to make predictions.
- They continue adjusting the bot in response to new data.
More specifically, machine learning is really a shorthand way of saying that quantitative scientists (also known as “quants) develop a learning bot for which they set algorithms—equations or rules for solving a problem—based on existing market trends. The algorithms used by AI get remarkably complex, but a simple example might look like this:
“Go long on stock X when stock X gaps up”
Once these rules are established, quants then feed that algorithm as much information about the market as possible. This includes information about stock prices but also relevant information about world events that impact those prices. For example, the knowledge that people might panic buy items such as toilet paper in the face of a natural disaster (e.g., a hurricane or pandemic) might lead an AI to make investment decisions on those items in the face of another triggering event. As they process the raw data, the bot distinguishes between relevant information, which it uses for predictions, and irrelevant information, which it discards.
The calibrations on these algorithms are remarkably minute and the margin for error is incredibly wide since a string of bad investment decisions can lead to a loss of clients who are seeking other options after losing their investments. Similarly, quants are needed to update the algorithm in the face of new criteria and information. The pandemic over the last few years is a perfect example of how new stimuli can generate ripples affecting multiple markets. It is for this reason that quant scientists are in such high demand in the financial sector.