Does the adoption of AI open up ‘flash crash’ trading exposure?
Does the adoption of AI open up ‘flash crash’ trading exposure?
As algorithms and artificial intelligence have been unleashed upon the world’s financial markets, so the term ‘flash crash’ has entered the lexicon. The original flash crash occurred in May 2010 when the main share indices in the US all endured staggering gyrations in the space of just a few minutes.
Such was the impact that the Dow plunged almost 1,000 points, or nearly 9%, wiping out over $1trn of stock market value before almost immediately recovering most of its losses. For those working in the financial markets, the incident was an earthquake where everything was immediately re-set as if nothing had happened.
Yet what is truly remarkable is that for all the publicity surrounding the crash, no one has yet worked out what actually happened.
The US authorities subsequently laid charges at the door of a day-trader living in west London, Navinder Singh Sarao, accusing him of an illegal activity called spoofing. But market commentators suggested subsequently that to blame the crash on one man was wrong and a subsequent report has suggested that high-frequency traders were not the cause of the instability.
Indeed, to this day no one knows the exact cause of the crash other than that algorithms are somehow implicated. As Yuval Noah Harari writes in her recent book Homo Deus, “experts have been trying ever since (May 2010) to understand what happened in the so-called flash crash. They know algorithms were to blame, but are still not sure what went wrong.”
It could happen here
So, to what extent is the betting industry vulnerable to similar extreme volatility in the near future? The nature of the trading that takes place between sports-betting and financials contains some similarities but also some crucial differences, explains Paolo Personeni, managing director of managed trading services (MTS) at Betradar.
“The average duration of a positon in sports-betting – that is the time to expiry weighted by cash flows - typically fluctuates between a minimum of few minutes and a maximum of perhaps a week, if we disregard outrights,” he says.
In financial markets, by contrast, it is measured in months and years, if not in perpetuity. “Hence, given that a sportsbook turns around much faster, the damages of a flash crash are much smaller in size, but much more difficult to recover before the exposure expiry.”
Still, the underlying technology employed in financial trading is much the same in sports betting, and certainly the widespread use of algorithms underpins every sportsbook today. Automation is already here and some are pondering whether a fully-automated trading desk of the type that dominate the financial markets is all that far away.
“Automation has existed in the sports-betting industry for many years,” says Personeni. “From event creation to resulting the process is largely automated for the bookmakers. We have been using clever algorithms in our pre-match and live-odds construction for many years. Nevertheless, the industry intelligence and intuition of an experienced trader remains key when preventing problematic outliers and corner cases.”
In financials, the fear on the part of the participants is that markets driven by artificial intelligence will soon take human agency out of the equation altogether but Personeni points to the difference between the currently deployed algorithms and AI when it comes to sports betting.
“Algorithms are used to replace complex processes or procedures with easy alternatives, potentially adding significant scalability to the business,” he points out. “But they don’t learn from their previous applications. AI, on the other hand, looks into the use of assisted learning and self-learning methods which correct or advance for future applications dependent on the previous outcomes.”
Still the potential offered by the evolution of AI is opening up potential openings in many areas of sports-betting and one startup that is hoping to make its mark is Statagem. It’s founder Andreas Koukorinis says AI-based products are sure to have an impact as machine-learning techniques are employed by trading desks to improve margins. But he insists that consumers will also see a benefit.
“The real benefits for the consumer lie in the entertainment value machine-learning technologies can bring. The introduction of advanced statistical analysis, and the ability to use computer vision gives users an unparalleled level of insight and boosts engagement with the platform they’re using.”
As with financial markets, the further advances in automated trading will enable reduced costs and higher operator margins. Personeni points to the extent of operations that now need very little real-time intervention, including liabilities, bet limits, market factors, customer confidence and event- and tournament-level confidence.
“Gone are the days of ‘referral to trader’,” he says of a platform where experienced risk and trading personnel sit alongside data scientists to configure the system to make decisions and provide optimised margin and exposure management.
Next up for automation is the risk-management process, customer profiling and clustering, automatic risk assessment and liability management. “But this will be evolution, not revolution,” he says. “It will still take time for the industry to both adapt and adopt.”
The human hand
Christopher Langeland, managing director at iBetting Cloud, the sportsbook backend business being developed by Gaming Innovation Group, tends to agree that for now human intervention will still be needed for the last mile of trading, much as it is in other aspects of the evolution of automation.
“We have been working with different forms of automation for many years now and I think we understand really well now what tasks can be automated and which ones still need a human hand to steer,” he says.
The central issue is that sports by their nature are chaotic and unpredictable. “An algorithm can’t be programmed that could anticipate absolutely every extreme scenario,” Langeland adds. “The day when we can leave the algorithms to run everything won’t arrive in my lifetime, I’m pretty sure.”
The technology may have changed but the task for the bookmakers remains the same – driving efficiency and mitigating against risk. “The speed with which sportsbooks identify the risk and react to the issue will be determined by the strength of their technology and data science approaches,” says Personeni. “The better we become at identifying the problem, the more proactive we can be by installing precognitive systems to prevent overexposure. The industry in its nature is very reactive; our goal is to be much more proactive.”
Much speculation over the near future centres on what jobs will be at risk come the next age of the algorithm. Whether it will take a lifetime for sportsbook traders to be under the threat of extinction is open for debate but the age of the robots is already upon us – the question is about how much more they might offer.