First quarter earnings season has been a resolutely downbeat affair this year, with many companies, such as General Motors Co. and Facebook, disappointing Wall Street’s expectations. The focus, as ever, has been on each company’s earnings release day and the very short-term market moves that follow each announcement. But what happens over the subsequent weeks?
There are many analytics tools out there catering to investors with an abbreviated time horizon, looking to make quick, next-day or next-week money off an earnings beat or miss. But for fundamental investors with a long-term thesis on a company, understanding how to modify positions going into and out of earnings announcements “has always been a difficult process to manage,” says Charles Trafton, a portfolio manager at FlowPoint Capital Partners, a Boston-based hedge fund firm.
EidoSearch, a Toronto-based startup, is hoping to change that. The firm’s founders, David Kedmey and Xiao-Ping (Steven) Zhang, met as MBA students at the University of Chicago in the late 2000s. Zhang, who holds a Ph.D. in engineering, is the technical brains behind the operation, while Kedmey, whose resumé includes a stint on the New York mergers and acquisitions desk at Oppenheimer, supplies the market perspective. Together, they’ve built a technical analysis engine that allows investors to understand assets’ future price movements based on historical patterns.
Most data analytics firms “have built algorithms that can search one page of text,” says Jeff Parker, EidoSearch’s chairman. “To search the whole web you need an engine.” Parker, who was No. 40 on the 2014 Institutional Investor Tech 50 ranking of financial technology’s most influential figures, founded pioneering equities research firms First Call Corp. and Street Events in the 1980s and 1990s. The industry veteran says the EidoSearch research tool lets investors “see things they’ve never been able to see before.”
For fundamental managers, EidoSearch’s technology is proving especially useful. It allows them to grasp the likely range of a stock’s future price movements based on both that stock’s performance into the earnings call and the different outcomes relative to analyst expectations (meets, beats, misses). This gives fundamental managers “a unique perspective into how to adjust position size in response to the probabilities,” says Kedmey. “How much should they hold? How much risk should they take off the table?”
Users select a stock and time horizons for both the backtest and the forward projection, and the EidoSearch engine combs the entire history of U.S. public equities to find similar pattern matches. It then provides a forecast range based on how the historical matches fared in the period following the test date. “I never come in in the morning and say, ‘Guys, we’re shorting the market — EidoSearch is negative,’” explains Trafton, adding that Flowpoint does not use the tool to pick stocks. “We use it to refine a position into and out of events once we’ve already made up our mind on what that position should be.”
EidoSearch now accounts for eight of the 40 variables in FlowPoint’s main market model. Trafton says that the tool, “like all technology, comes with its own biases, but it’s very useful when you find those outliers where seven out of ten times the evidence is obviously bullish or bearish.” Digging into the history of the matches allows investors to understand why their approach may be wrong or in need of adjustment; it gives them context to reconsider their take on present events.
Predictive analytics tools have enjoyed something of a mixed record since the financial industry erupted with excitable talk of the possibilities of “big data” a couple of years ago. Most of the products pushed into the market in that first flush of enthusiasm relied on historical numerical or verbal data — much of the latter culled from the Internet, Twitter in particular — to generate signals and predictions around asset price movements. The usefulness of these products to investors is a topic of much debate; traders see dwindling value in the type of “sentiment analysis” that crawls social media sites to understand the likely movement of stocks, for example. The difference with EidoSearch, which has received $2 million in seed funding from San Francisco venture capital firm Resolute Ventures, is that the technology is geared toward finding visual pattern analogs — an unusual and difficult task.
Kedmey and Zhang took more than five years to build the technology underlying the EidoSearch engine; the commercial version of the software was released late last year and the firm now has 50 buy-side customers. “We’ve tried almost every fintech platform in the last 20 years,” says Trafton. “A lot of them are just noise — firms start with a bang, make a lot of noise, then fizzle.” But the EidoSearch tool is “very elegant technology,” he adds. Kedmey and Zhang will be hoping more investors feel the same way as EidoSearch seeks to avoid becoming another predictive analytics fizzler.
Follow Aaron Timms on Twitter at @aarontimms.
