Learning Curve: new data is changing markets for good
New regulations have uncovered heaps of data that markets are struggling to handle. But as participants discover issues exist they did not realise existed, the data revolution will result in safer and more efficient markets for all.
By Alexander Lamb, head of marketing at The
Knowledge is power and new regulations like MiFID II have revealed a lot of very valuable, untapped data. Data science, previously almost exclusively the domain of retail businesses, is now in the premier league of finance.
This has resulted in better connected and understood market structures allowing for better risk management and market monitoring, not to mention benefits to the business.
But the levels of complexity are greater than most imagined possible and studying them in isolation just does not make much sense.
Then and now
Prior to the Markets in Financial Instruments Directive, very little was known about who was doing what, apart from rudimentary data on trades and end-of-day statements.
At best, data was accurate. But it was often formatted in hundreds of different ways, submitted in an untimely fashion, and was often incomplete.
At the start of the new era of monitoring, erroneous or missed alerts abounded. Now, however, the ability to spot the proverbial needle in the haystack as soon as it is present is no longer a dream: it is becoming a reality.
We can now closely watch orders and trades across their lifecycles. We have also learned that when data is delivered to monitoring systems influences how we perceive their risk. For example, what if two legs of a strategy trade — for example, selling a cash US T-Bond and buying US T-Bond Futures microseconds later — arrive out of sequence while the others did not? If the cash bond leg is received after the futures purchase the net exposure in face value terms may look like a limit breach — long too many cash bonds and long too many futures!
New approaches to storing and retrieving data are being used. Traditionally organised, pre-sorted, pre-specified data can limit discovery. Neatly organising transactions into class, instrument type and maturity silos means that finding links between these and other activities requires a lot of planned research. The user has to look for patterns within or across these silos, for example.
The new approach of having ‘NoSQL’ (meaning either ‘No Sequel’ or ‘Not Relational Sequel’) databases means flexible databases have
Knowledge from mining data, both in real time and from these flexible data stores, is a race against time that every business manager and risk manager wants to win. Hardware technology continues to improve as well, making the impossible possible.
Managing ‘unseen risk’
It is beginning to hit home that we underestimate the potential volumes of data that the now almost fully electronic or automated trading world can deliver. As trading systems from exchanges down to the decision-making platforms become increasingly reliable, they can handle greater volumes.
As transactions contain more information, the growth will continue to be exponential. No single firm can predict these changes but must anticipate them, or miss the next opportunity.
The trader manager can now see many activities in an independent way. There is no longer a need to employ a programmer to deconstruct an algorithm that drives a trading strategy to understand how it works — just diagnose the behaviour.
Knowing how behaviour changes from known to new patterns can be strong indicators of rising and so far unseen risk. This would be the same for risk managers and clearers (FCM/Brokers).
The trader who normally closes positions at a certain time of the day, but suddenly starts to increase those positions, could be having a good day and be getting very bold. The trader may still be within his or her limits, but the new activity should ring some alarm bells.
Transaction data can be considered in context with the markets — the best execution rules have brought market data into the live monitoring arena. Risk rules are also looking at more than simply the order at the time it was submitted, changed or executed.
These rules are typically configured in line with limit settings in trading systems, for example with maximum long or short values, what markets are allowed and maximum order sizes. Seeing how these limits are tested is raw data waiting to be analysed.
Limits may currently be arbitrarily assigned based on outmoded criteria — risk measures can now be seen in real time — so it is now possible and desirable to set limits in real time too.
‘3-D’ market view
Riskier transaction management can lead to extremely risky positions in milliseconds or microseconds. Identifying those and alerting the responsible people is beginning to be the standard that all sides expect.
Equally, identifying consistent good behaviour that is over restricted by conservative rules may be an opportunity for increasing limits when the limit check factors or values are all within acceptable parameters. These factors include maximum positions and intraday profit and loss (P&L) limits for example.
That means even assessing market conditions and liquidity relative to the positions held and the orders in the markets, could positively influence the risk parameters, allowing more trading.
The earliest approaches to monitoring limited us to a relatively two-dimensional view of activity. We now have a three-dimensional view, where events are no longer isolated and we can truly analyse trading behaviour.
No more isolation
The increasing importance of data has also resulted in more pooling of information, with firms increasingly working together. No two firms have similar clients. An over-the-counter market is not displayed in its entirety, nor is a hybrid voice and exchange market (US Treasury Bonds for example).
Trying to fill the gaps in isolation, whether you are a trading or clearing firm, a vendor or an exchange will soon be outdated.
This ‘sharing’ has begun with mandatory reporting of trades to a central repository. The visibility of these trades to the manager responsible for them could lead to and is probably already starting conversations about how these are perceived from a risk point of view.
Regulators pass on comments and concerns in their questions while firms talk to their analysts and vendors about these new developments.
These break down barriers without giving away any trade secrets. Industry associations and vendors constantly examine and discuss both the business and risk concerns, with innovative tools for examining the key points of interest being created to test theories or concerns.
Vendors are aware that they cannot cover everything and actively seek out partners who will contribute significant and important pieces of the information gathering and monitoring services that they cannot achieve individually.
Sharing more non-proprietary information, developing a better understanding of the different markets and participants, instruments and strategies is an essential part of this new data-driven world, in which we will be looking for new trades and improved ways of ensuring their safety.
Clearly, there is more to learn, but we now have a lot more information, tools and, not least, partners with which to share and examine that data and create safer markets.