The exponential increase in trading data has become a challenge for traditional commodities venues who “have come late into the electronic age”, according to Bob Mudhar, partner at Citihub Consulting, while the best solution to their issues will be found in artificial intelligence (AI) and machine learning, and not in the cloud.
“As trading becomes more electronic, the volumes of market data, orders and trades also increase. Managing this requires investment in newer storage, event handling, and data analytics solutions,” says Mudhar. “Since much of this data is also publicly available, venues also run the risk of falling behind other market participants in understanding the markets they are supporting. Venues that have traded commodity derivatives have tended to be addressing increasing data volumes for much longer than physical venues.”
Mudhar says there is little benefit in the cloud, as the software used in commodities is “generally proprietary, and on-premise for security and reliability reasons.” There is little reason, he adds, for firms to build a cloud-native version of their software, as the benefits which the cloud traditionally brings “do not sufficiently apply” to commodities. “The venues are doing risk management, but this does not yet demand the same scale of which investment banks are facing.”
Despite this, there could be a “second wave” of adoption for trading venues when they ride the wave of cloud-based data analytics. “The advantages are that you can crunch significant volumes of data using well engineered PaaS solutions, and they have little in the way of existing solution so there is a data analytics ‘green field’. We don’t think that the volume of data itself will preclude use of the cloud from a cost perspective, and the cost will be in very ‘peaky’ analysis activity which lends itself well to the cloud pricing models.”
75% of respondents to an August 2018 survey from FIS and the Commodity Technology Advisory reported that cloud was an area in which the greatest amount o of investment was being spent. The report predicted that revenues from cloud-based CTRM systems would grow between 10-15% each year to 2021. Cloud, it argued, is an effective entry point for smaller firms “seeking the lowest price options possible”.
When it comes to automated trading, Mudhar states that “equity, fixed income, and FX markets have had algorithmic trading engines operating on regulated markets and MTFs for many years, and for equity markets, for well over a decade. These engines have some limited decision-making discretion within well-defined trading parameters. If AI is the use of rules-based logic to make decisions more reliable and predictable than a manual version, then in that sense these markets have AI and some level of machine learning.”
“AI and machine learning can automate tasks to reduce risks and increase profitability,” said Dr Daniel Bloch, head of machine learning at Brainpool.ai, in an email. “The commodities market is very sensitive to geopolitical risk, one of the biggest efficiencies that AI can bring is the ability to quickly analyse a huge amount of data around the risks – using natural language processing and sentiment analysis – and then provide more accurate forecasts about how those risks impact the market.
Mudhar says it is important to remember that trading venues compete for transaction fees with each other. “Venues are paid by transaction volume – they have fees for exchange, clearing, settlement, compression, etc. So, we can expect the electronic venues to want to support high-frequency trading (HFT) so they can increase their fee volumes. HFT firms care little for the asset being traded. It’s all about apply their strategy to a market. Venues want that volume.”
The US Commodities and Futures Trading Commission (CFTC) released an industry analysis in March 2017, reporting that more than 60% of crude oil trades and more than 50% of metals trades are being performed by automated systems. In 2015 the regulator proposed a series of controls on automated trades to avoid “flash crashes” in the market caused by largescale automated orders. Named Regulated Automated Trading (Reg AT), the proposal was subject to a formal consultation in earl 2016, but no timescale for the adoption of the rules has been set.
“Just because the CFTC has not acted on an algorithmic or automated trading regulation does not mean that we are not focused on the impact of such activity on the market,” said commissioner Brian Quintenz, in a June 2018 speech.
In Europe, the second Markets in Financial Instruments Directive (Mifid II) includes rules for entities in algorithmic and HFT trading. Firms must comply with numerous requirements, including: notifying its national regulator that it is involved in this type of trading, potentially disclosing its trading “source code” to its national regulator, and maintaining particular risk controls that are subject to certain testing requirements.
“Making better use of their existing data is going to be critical to commodity markets in the future. This will enable the markets to make the trading of their own members more effective," says Mudhar. "Regulation will turn to commodity markets soon enough. We started to see this with Mifid II commodity position limits reporting. It will not be too many more years before Mifid scope extends beyond ‘financial instruments’ to include FX and then commodities.”