By Adam Jones, Sales Director, UK
For many years, financial markets have remained relatively stable. Despite events as severe as the global financial crash in 2008, our markets were very efficient, our interest rates have been very low, and on the whole volatility has also been fairly minimal. Recent events, however, have led to quite a different situation. One that few of us could claim to fully comprehend. While the future is perhaps difficult to predict, the path that led us here is very clear. Here, we’ll dig a little deeper in to that journey and look at what that might mean when the worst of the COVID-19 pandemic is over.
With high market volatility a thing of the past, the opportunity to make significant returns on relatively small deals is much rarer than it used to be. Old playbooks are no longer valid and new approaches are needed.
Most financial services businesses have long realised this and are taking a different approach; one led increasingly by technology and relying less on gut instinct and relationships of traders.
As technology advances, financial services firms find ways to use it to their advantage. Utilisation of new technology has not been spread evenly across the industry, nor applied in the same way by all participants. There is huge diversity in how organisations. From savvy start-ups to the largest and most well-known global financial institutions, firms are deploying technology to their advantage.
There are two very popular approaches that many institutions are following to certain degrees, but are by no means exclusive – in fact, the most successful firms are arguably those doing both of these things.
The first is trading faster, capitalising on microvolatility. While the overall yields on individual trades may only be slightly higher than a more traditional approach, the ability to execute more trades orders of magnitude higher than previously in a given day amplifies the gains; meaning these shorter-term bets generate larger overall returns.
The second is trading smarter, capitalising on intelligence. Today there is a wealth of data available to traders. With markets more automated than ever, some would say the movements are more possible to predict. The basic principle is that data is collected, processed, and analysed by powerful computers to generate the intelligent insights used by traders to increase their win rate and the size of those wins.
The rise of algorithmic (‘algo’) or quantitative (‘quant’) trading has been expontential in recent years, as more and more firms are turning to computers and machine learning alongside their MBA graduates to bring in their revenues. Even within quantitative trading there are several different approaches – the data that enters the model being key.
As the old adage goes; ‘garbage in, garbage out’. Even if your model is perfect, if you feed it bad data, you will never get a good outcome. Some firms have been incredibly innovative in selecting just what data is fed in to their model. For example, Quandl tracks the flight details of private jets, using intelligence to predict how corporate executive travel is likely to impact the markets. Combining this with other data sets and intelligence can yield significant returns.
In 2019, a Gulfstream jet owned by Occidental Petroleum flew in to Omaha airport in the US. Quandl spotted this event and quickly alerted their hedge fund clients of the unusual visit and a potential trading opportunity. As it so happens, Occidental had sent a delegation to meet with Berkshire Hathaway chairman Warren Buffett, hoping to persuade him to help fund their takeover bid for rival, Anadarko. Two days later, Occidental announced a $10bn investment from Buffett. That 48-hour head start enabled Quandl clients to win big when news of the deal broke.
This new frontier in the Information Age is one that many are finding promising. The returns from the digital ‘Gold Rush’ may be higher, with the ability to lean more heavily on technology to support decision-making. Increasingly intelligent, machine learning enhanced business models, require increasingly intelligent systems to power them.
Pooling large data sets and applying complex machine learning and deep learning models to better understand how markets and their makers behave is a complex technology challenge and one that’s not to be taken lightly.
The Hardware Challenge
The infrastructure needed to build and train models to trade quicker or smarter, or both, is significant – and not just in size and scale. Highly specialised equipment is often needed, which mustbe highly secure and operated to national and international compliance standards. Bankers cannot build these systems in their basements. These workloads need to run in datacentres with much higher levels of power availability, connectivity and compute power than can be run on-premises in an office building. To build the kind of quantitative models we’re talking about in an efficient way, hyperscale datacentres equipped with high performance GPUs, and capable of handling very high power densities are required.
Once machine learning models have been built and trained, they can be run closer to key liquidity venues, with low latency access to key exchanges and market participants. In London, our data centre campus is just a short walk away from most City offices and is by far the most connected location in the centre of the city. This means that ‘inference’, or the deliver of the output from running these models, can be delivered in milliseconds to those who need it.
Data Centre Innovation
The kind of remote model building discussed above is becoming more and more common. And with it the role of the data centre in financial services increases in importance. Being able to locate enhanced data processing capability nearer to trading venues is becoming easier thanks to colocation services provider, like Interxion. Partnerships with chipmakers such as NVIDIA to deploy their specialised AI chipsets in sites supporting the financial services community make it easy to deploy sophisticated market models and intelligence algorithms where they are most effective.
Deployment of dedicated AI appliances in colocation data centres enhances traders’ ability to crunch data and deliver results of that analysis on the timely basis. Traders can identify the investment opportunities and make greater returns than their competitors.
Increasingly, financial services choice of data centre partner and their use of their services, is therefore a critical factor in their success. CIOs know that they need to adapt to stay afloat and ahead of their competitors now that the market is more fluid.
With the COVID-19 crisis providing many of us with a chance to reset and reflect on our strategies, financial services will continue apace with digital transformation and new trading strategies as we emerge from lockdown. I have no doubt that when the markets do settle and an efficient, low volatility world returns, those with better machine learning models will be the ones outperforming the competition.