2 min read
Identifying hard-to-borrow securities just became easier.
By Eugene Kanevsky, Global Head of Electronic Trading, CLSA
The concept of artificial intelligence (AI) conjures up visions of a future world yet AI already influences us in our daily lives and has done for some time. From simple things such as suggesting friends on Facebook or proposing our next Amazon purchase to delivering news of interest direct to our phones and even through the apps that guide us by analysing real-time traffic conditions to suggest faster routes.
The financial markets are also changing and algorithmic trading, previously seen as the leading edge technology, is now seen as static or even outdated unless it is augmented with AI technology.
At CLSA we have launched the next generation of algorithms based on our proprietary AI machine learning framework, called “ADAPTIVE”, reflecting the ability to continually adapt in real time to the market.
Recently we have heard a lot about the use of AI, predominantly with regard to stock classification initiatives. In effect, historic stock trading patterns have been analysed and classified into similar groups using machine learning techniques. These groupings then assist traders in selecting more appropriate trading strategies.
AI models essentially redefine and expand upon the traditional groupings traders understood and used daily, such as large- or small-cap, liquid or illiquid and wide or narrow spreads. These models based upon stock classification are developed and continually refined to reflect trading pattern changes.
The objective of AI stock classification is to optimise algorithm selection for stocks, and while the development offers some benefits, it does not address the fundamentals of the algorithms used to trade in the market.
At CLSA, we wanted to know: “how will the stock trade next?” We knew that by answering this question correctly and regularly with a high degree of certainty would provide far more value to our clients’ execution.
We base the CLSA’s ADAPTIVE platform upon a proprietary neural network.
Neural networks are loosely inspired by neuroscience and the most effective machine learning method known today. This trained neural network generates real-time signals projecting what is going to happen next in the market enabling the ADAPTIVE algorithm strategies to adjust execution plans as they trade, rather than rely on static predefined scenarios.
Our neural network is currently trained to predict short-term and long-term price movements, volatility and trading volume, but this is just the beginning in a process of continual development.
In recent years, we’ve seen a growing number of non-traditional market participants exerting their influence on the Asian trading landscape, in some markets reportedly accounting for half if not more of the total traded volume. Much of this liquidity is automatically created by various types of quant algorithms.
Algorithm strategies need to be smart enough for this new world and that’s where the implementation of effective AI proves invaluable.
Algorithm trading has always been an arms race. For many years that arms race was largely limited to brokers with their algorithm quant and development teams releasing the next iteration of their latest, greatest and smartest enhancement to their suites of strategies, be that quarterly or annually.
However, equipped with nimble, shorter release cycles and increasing utilisation of AI technology for prediction, it’s not hard to understand why short-term horizon non-traditional trading houses have found short-term profitable opportunities in many Asian markets at the expense of long-term investors.
In response, brokers have grown the number of quants and technologists in their algorithmic development teams in an attempt to shorten release cycles and keep up with the trading landscape changing at an ever increasing rate.
The man-power versus automation scenario has played out in many industries, but it’s always more striking when it’s in your own area of expertise. Now, with the benefit of hindsight, observing ADAPTIVE do its work, taking just minutes to detect new patterns, multiple times during the trading day and at different price levels, the futility of more people working harder against smarter technology becomes apparent.
CLSA’s ADAPTIVE short term price prediction signals achieve an accuracy of over 95% in volatile securities. We have implemented this AI technology in our agency algorithms to provide clients direct access to this impressive technology.
Earlier I mentioned the challenges of competing with non-traditional market participants such as HFT and other short horizon strategies. While identifying short-term price displacements is an important element when improving execution, the ADAPTIVE framework also predicts future trends, volume, volatility and price, allowing the agency algorithms to optimise execution for the more traditional agency trading benchmarks, such as VWAP, IS, Inline or Close. Results have been impressive, and we see a win-ratio when compared with traditional algorithms in excess of 80%.
One of the common misconceptions is that this new generation of agency algorithms with embedded AI are black boxes, and the sales traders’ role will be merely to enter the client order and then let the technology do its work. We do not agree. AI algorithms are there as an empowering tool for the sales trader and dealer to use within the overall trading plan.
Buy-side traders, with their knowledge of the investment objective and fundamentals, are essential in mapping out the appropriate trading strategy. The sell-side trader brings detailed knowledge of the market and current sentiment: AI algorithms provide an edge in the marketplace.
CLSA’s interpretation of best execution is the sum of all the parts. We continue to believe in local market knowledge, the value of sourcing block liquidity and the importance of delivering timely and relevant information to our clients. Leading the way in market execution technology in Asia is a critical component.
We’d love to hear your feedback on this article. Please click here
4 min read
AI and machine learning can enable traders to manage trading risks better and enhance investor returns. By Ben Polidore, Managing Director,...