Who’s Afraid of the Big, Bad Algorithm?

3 minutes   |  SYSTEMS & TECHNOLOGY

The views expressed in this column are those of Nir Vulkan, and may not reflect those of Saïd Business School, University of Oxford, its faculty or GetSmarter, a brand of 2U, Inc.

Robots. If they are not crashing into us in their self-driving cars, then they are plotting to put us out of work. Apparently, no one and no job is safe – and now even our savings are apparently under threat. Recently, the Daily Mail asked its readers, “Would you trust a robot to manage your pension?” The implication being – are you crazy?

Well, I’ve been building algorithmic trading models for over 15 years and teaching the theory behind it at Oxford University; my answer is, yes, I would. The scaremongering you find in the press plays partly on the feelings of inadequacy most of us have about looking after our finances “I really should be doing more about my pension” and partly on our general fears about faceless robot overlords.

AI algorithms: Past, present and future

Robots and artificial intelligence (AI) seem like a new threat, because computing power is increasing so fast, and because mobile phones and other devices are gathering and analysing our data like never before. We’ve been living with this technology for a long time. Software robots, known sometimes as ‘’bots’’ or simply, ‘’algorithms’’ have been around since the beginning of computers, and have been a routine part of the finance industry for more than thirty years.

The scaremongering you find in the press plays partly on the feelings of inadequacy most of us have about looking after our finances “I really should be doing more about my pension” and partly on our general fears about faceless robot overlords.

While new companies are creating new promising AI-based algorithms and amazing performances, the truth is that Algorithmic trading has been used by leading hedge funds since the 1980s, and the tech is now considered mainstream. While hedge funds are primarily for wealthy individuals and large institutional investors, you and I may also have had some exposure to them, though our pensions. Some of these hedge funds, sometimes referred to as “systematic funds” have done so well over the years that they now manage billions of dollars.

At its best, algorithmic trading provides automation to a process that takes out the emotional response to sudden market moves, or anything else which might scare or excite humans. By using an algorithm to trade, we are forced to commit to a set of rules, and only trade when the right opportunity presents itself.

For example, some of these algorithms trade what we call “momentum” – that is, they see strong market moves, like a surfer spotting a big fat wave, and they speculate that it will take some time before this wave breaks. That means they can take the correct position and wait until another mathematical indicator signals that the wave is about to come down. To switch the analogy from surfing to flying, it’s a little bit like having an automatic pilot to analyse all the available information to get you to your destination, smoothly and safely.

Furthermore, since algorithms use such powerful computers, they can facilitate types of trades that humans would find very hard or expensive in terms of manpower to do. For example, a computer can trade a model in 100 markets at the same time; or it can analyse data very quickly and make a decision every few seconds, which it will update again when new information comes in.

Using AI algorithms — the good, the bad, and the ugly

Is it all good? Of course not – there are bad algorithms too. In my experience, these fall into two categories – ones that deliberately try to manipulate markets (like the one behind the flash crash of last summer, where the US authorities are now bringing prosecution against a West Londoner who used a simple, but effective algorithm to cause share prices to dramatically fall for few minutes causing all kind of mayhem in the financial markets), and the ones that were designed in good faith, but failed to put in place proper risk management procedures.

There will always be people, in any area of life, who try to manipulate and cheat. All we can do is make sure that such activity is illegal, difficult to achieve, and easily traceable after the fact. While computers may offer fraudsters the ability to act more quickly, it also means their crimes can be detected and punished, so that others will be deterred.

What about the second category – accidental harm? My advice is to make sure you have at least one person in your team with experience in real-life trading and risk management. Trading is ultimately about managing risk, and not about outsmarting everyone else. The collective “smartness” in the market is smarter than any of us individually. I have seen many clever mathematicians fail in this arena, because they did not think enough about trading costs, slippage and what to do in a drawdown. If you’ve ticked all those boxes, then your algorithm will do whatever it is you wanted it to do.

Nowadays, I spend more time talking to engineers, computer scientists and physicists than I do to fund managers — All looking for a piece of the action, claiming to have better, smarter AI than anyone else. I’ll say to you what I’d say to them; trading is about risk management. Good algorithms focus on the opportunity and reduce exposure to all other risks. Bad algorithms? Well, these are the ones that even the people who wrote them don’t really understand. Avoid them.

Algorithmic trading systems must be designed properly by people who understand and respect financial markets. Like any other machine, AI algorithms can help us be better at our jobs and achieve our goals. So, the next time you read a scaremongering headline, switch the question around. Would you really trust a human to look after your money?

Algorithmic trading systems must be designed properly by people who understand and respect financial markets.

 

Professor Nir Vulkan

In addition to being an Associate Professor of Business Economics at Saïd Business School, University of Oxford, and Programme Convenor on the Oxford Algorithmic Trading Programme; he is a Co-Convenor of the Oxford Fintech Programme and the Oxford Blockchain Strategy Programme. He is also a leading authority on: e-commerce, market design, applied research, and hedge funds. Professor Vulkan has authored two books and over 40 articles in international journals in all aspects of technology, economics and finance.