Automated traders are in an elite club. Every day, they walk a high-stakes tightrope between the potential for big gains and big losses. Whether you’re a market maker, volatility trader, or systematic trader, every day is like a poker tournament held at microsecond speeds.
That’s why speed commands so much attention in the machine trading world. Speed is sexy, speed sizzles. In shops that have more of it, their machine-driven trading can move them into or out of positions faster than their competitors. That helps win more often than not, and equally important, helps them avoid getting picked off.
There’s no question about it. Speed is exhilarating. But it’s also stressful.
When you’re moving at microsecond speeds, very good and very bad things can happen in a flash. Therein lies the stress. You and your machines need to be always consuming data, sifting it for opportunities, and moving on them fast – and your moves need to be right most of the time.
It’s that last part that can often get tricky. With machine traders focusing so much on speed, they seem less focused on the quality of the data they’re feeding into their models – the ones and zeroes that ultimately drive trading decisions. That’s dangerous because bad data received at the speed of light is still bad data.
Take event-driven opportunities, for example. Lots of different types of events can create volatility and intraday trading opportunities. For most of them, getting the information quickly and understanding its significance faster than others is how traders gain an edge – so it’s all about speed. That goes for macroeconomic events, like change in the U.S.’s natural gas supplies, and unanticipated events, like news of a major airline CEO defending his company after a passenger is randomly chosen to be removed involuntarily.
But with planned or largely predictable events, the quality and accuracy of that data is even more important than how fast a trader gets it.
Many U.S. publicly traded companies announce their earnings dates in advance by publishing this information via press release or other means. Investors and vendors alike scour the content to get accurate information on the upcoming earnings release. Yet, with access to multiple press release services and a dizzying number of traditional and new venues (i.e. social media) to announce the upcoming date, there are many options for each company and a lot of bases to cover for investors/vendors. Without a thorough investigation of the company-issued data, some assume they can just use last year’s date plus one day. That’s not good enough and not at all good for a risk or trading strategy.
Further, what about the companies that do not make use of standard press releases or news stories to announce their upcoming earnings date? Instead, they may use regulatory filings, disclose the date somewhere within their corporate web site - either in writing or in other media (like video!) - or announce it at investor meetings, analyst days or on conference calls.
Many of those methods aren’t machine-friendly, so the human factor enters the equation. Sometimes people do a great job, knowing where to find all the necessary data, inputting it accurately and constantly updating and performing QA to their work. Whether a company is changing a date, revising a schedule or adding a speaker – capturing revisions to the event data immediately is critical. Sometimes the revisions aren’t caught, and when that happens, it’s easier for you to get picked off.
In addition, let’s not forget the importance of accurate historical data. That way when the signal arrives, you can act quickly knowing you have completed a clean backtest.
The bottom line for machine traders is this: All the processing and trading speed in the world won’t matter if you get it wrong because of inaccurate data. And there are relatively few places in the trading world where data is pure (untouched by human hands). It would be wise, therefore, that in addition to speed and performance concerns, that machine traders take a hard look at the quality of the data they are using and from where they are sourcing it.
What are the options? Some choose to ignore the issue and hope for the best. That has loss and disaster written all over it. The second option is to build a process in-house for fixing the data. This choice can work well, but often, it’s expensive and difficult to maintain—and constantly needing attention across all the disparate data providers. Option three is to partner with specialized data providers who are totally focused on their data offerings and highly motivated to get them right.
As with everything in life, when it comes to data providers and their quality, you get what you pay for.
Machine trading is like a refined, high-stakes tightrope walk, with more than its share of risk, volatility and stress. By making certain that your systems only operate with the best data available, you can reduce that stress and confidently outperform your competition time and again.
Barry Star is CEO of Wall Street Horizon.
Also published in Trader's Magazine.