';
Views Article – Sharenet Wealth

South Africa

Quantitative trading explained for people without a mathematics PhD – PART I

Part I: “Mean Reversion”

The world of quantitative and algorithmic trading is a complex one – sometimes unnecessarily so. Generally speaking, it’s an inaccessible one too: it requires some understanding of mathematics, programming proficiency, and the capital and tools to implement the strategies. The field is further complicated by terms understood only by those in the know such as mean reversion and statistical arbitrage. Is this a grand conspiracy to make the average Joe/Jane believe they are incapable of understanding quantitative trading? Perhaps; perhaps not. Regardless, the computer-driven techniques employed by hedge funds could assist you in your human-driven trading strategies too.

Many are under the impression that quantitative trading is one-part maths and two-parts black magic. In fact, any quantitative analyst (or simply “quant”) worth their salt employs the scientific method in the same way as any physicists, astronomer or statistician would (it’s no surprise that many quants originate from these fields).

Any algorithm begins as an observation of some phenomenon in the market – usually based on some intrinsic economic or fundamental principle and it’s usually simpler than one would expect. For example, Anglo American Platinum (AMS) and Northam Platinum (NHM) are both in the exact same business. As they are affected by the same factors (platinum price, exchange rate, policies), it’s reasonable to assume their share prices will move in tandem. There is an underlying relationship between the shares that quants seek to exploit. Every so often, one of the shares will become overvalued and the other will become undervalued – causing the relationship to be off balance. For the relationship to revert to normal, the overvalued share should theoretically drop and the overvalued should rise. By selling the overvalued share and buying the undervalued share, a neat profit can be made. Looking at the graph below, one can see that NHM is quite lower than AMS.

These concepts form the fundamental basis of pair-trading. Quants will take this observation further, research academic literature on the subject, backtest a hypothesis, and determine whether the strategy is profitable. Once its profitability has been established, quants will go further and fully automate the strategy. 

In the past decade, AI and machine learning has come to the fore and opened a whole new branch of quantitative trading. However, classically there have been two main schools of thought: mean reversion and momentum-based strategies. Pair-trading (such as the AMS/NHM example) is a mean reversion strategy. Mean reversion is a fancy term that essentially says there is some phenomenon, such as the relationship between AMS/NHM, that can be predicted and exploited once that phenomenon moves out of the ordinary. Looking at the graphic, the phenomenon moves up-and-down around some average. Sometimes it moves up a bit much and sometimes it drops too much. After every extreme move, it can be predicted with high probability in what direction the phenomenon may move next. That’s when profits can be made. 

One needs to understand that financial markets are random. Very, very random. And most trading strategies are shown to be unprofitable once costs are considered. However, pair-trading is a widely successful strategy even after fees and, as it turns out, AMS/NHM is statistically one of the best pairs to trade. One doesn’t need an algorithm and automated program to trade it. Sharenet’s TraderGo has a Z-Score indicator which works like the RSI. When the Z-Score much higher/lower than 0, it is expected that it would revert to 0 (hence, mean reversion). 

At the date shown by the speech boxes, the Z-Score is very high – indicating that the first share (AMS) is overvalued relative to the second share (NHM). To trade the pair, AMS is sold and NHM is bought. The trades will be closed when the Z-score gets close to 0. Granted, NHM’s price decreased slightly and since it was bought a small loss was incurred. However, AMS was sold and its price decreased significantly making a tidy profit overall (the profit on the AMS trade offset the loss on the NHM trade). 
 

Pair-trading is a classic quant tool. There are thousands of possible pair combinations and not all are profitable. What discerns a good pair from a bad pair? Well, without statistical tests its not possible to say for certain. However, provided one can justify some fundamental link between two shares (such as a shared industry), the pair has a reasonable chance to be a good one. You don’t need a PhD to know that Coca-Cola and Pepsi are bound to be related.

Like this article? Get more like it and exclusive market alerts before the general public by subscribing to free Sharenet newsletters using the form below.

Register

We want to get to know you

* Required Fields

Aidan van Niekerk

Junior Investment Analyst

Aidan is responsible for quantitative research and stock-market analytics. He has a keen interest in all things numbers and - in addition to working for Sharenet -  is a full time BSc Mathematics and Statistics student at Unisa. Aidan has represented South Africa at a professional level in road cycling where the rigorous demands and self-discipline has prepared him for the markets. He aspires to one day complete a Master's degree in statistics and delve deeper into the world of algorithmic trading, market analytics, and financial modelling.