CAN A ROBOT GO CRAZY?

We have seen it in science fiction countless times. The robots get smart, get angry and then rise up against us. Or some glitch results in widespread chaos with planes falling from the sky and microwaves exploding. For investors who entrust their money to robots this is always a concern lingering in the back of their mind. But when it comes to Artificial Intelligence investing, can this ever actually happen? Let’s talk a bit about robots and the technology that stands behind them.

WHAT IS A TRADING ROBOT?

A trading robot is not a computer or an android like R2D2. A trading robot is a program or an algorithm. It is programmed to receive data, for example – market prices, written analysis or even posts on twitter. Once the data is received it is run through a set of equations which are designed to indicate if the current market represents a trading opportunity. If there is an opportunity the algorithm generates a signal to buy or sell one or more financial assets. As a rule, such robots or algorithms tend to be quite simple, quite volatile in their performance and they tend to deteriorate over time as market behavior changes.

To combat this, the trick is to combine many of these robots together into a balanced portfolio. This method is called an ensemble, i.e., robots are united into an ensemble, and they trade together.

ENSEMBLE OF TRADING ROBOTS

Ensembles of trading robots increase portfolio profitability and reduce volatility by spreading the risk and allocating to the best performing robots. By combining lots of different type of robot together the overall model performs well in different market conditions and produces a much smoother return schedule.

Robots are currently managing $ 1.6 trillion out of the total $ 80 trillion assets under management globally. More than $ 9 trillion of assets are expected to be algorithmically managed by 2025.

HOW IT WORKS

Imagine a symphony orchestra playing Vivaldi’s Four Seasons. If only a trombonist turns up to the concert, for example, the piece will not work. But if all the musicians come, every note is played, and the music sounds fantastic!

Many people spend their energy trying to build the perfect robot. Don’t bother, constantly performing robots do not exist. There is no holy grail. Robots are expected to deteriorate, misfire and eventually get replaced.

All robots in the ensemble work independently from each other, which means they do not know about each other’s existence, do not engage with other robots, and, accordingly, do not receive commands from their comrades.

As a result, we can accept a certain level of failure within the group. In fact, we expect it. But because there are thousands of robots, if one experiences a malfunction, it does not have a major impact on the group. The offender is simply removed and replaced with a new version.

IMPORTANT: we do not consider a market drawdown as a “failure” or malfunction. Robots are expected to both make and lose money. The aim of the ensemble is to collectively generate a profit more often than it generates a loss. This is where the impact on the Sharpe ratio is so important. But by engineering Sharpe higher with an effective ensemble we can significantly reduce the risk of loss. For example, with Sharpe = 1, the probability of generating a profit over one year is around 84%, with Sharpe = 2 it’s 97.7%, and with Sharpe = 3 it’s 99.9%.

The threat of AI Armageddon will remain in science fiction. But in reality, robots are small, replaceable and too limited to be a danger to you or your money.

Access for free ETF Quant portfolios

arty ETF portfolios are based on AI