February 17, 2014

Using sophisticated pattern recognition technology, Eidosearch can provide equity traders, analysts and portfolio managers, with the ability to project return expectations for thousands of stocks, based on decades of history -- all within seconds, notes its web site.

"We offer the ability to uncover the recurring patterns throughout time, in a very fast and accurate exercise," says Kedmey, who previously worked in the mergers and acquisitions department in the investment banking division at Oppenheimer & Co. He cofounded the company with Xaio-Ping (Steven) Zhang, whom Kedmey met at the University of Chicago Booth School of Business, and is expert in information processing technologies and content search engine.

Zhang, who is CEO of Eidosearch, has applied content-based search to photo matching and the human genome project. "When we met at the University of Chicago we realized that these search techniques were well suited for financial data," says Kedmey. Unlike looking for data in Youtube and in social media, in finance the data is well defined and well structured," said Kedmey. Also, if someone wanted to describe the behavior of the VIX and picture the curve and how it trades over time, language is not well suited to describing that path, says Kedmey. "With patterns recognition you can search with patterns, there really is no semantic gap," he explains.

Ultimately, Eidosearch is trying to quantify investor behavior. For example, last year, investors shot Tesla Motors from $40 in early April up to $193 by the end of September. In response to reports of three battery fires in early October, investors became fearful and began to bail out of the stock, producing a chain reaction that pushed the stock down to $121 by late November. Though Tesla's fundamentals did not justify a drop of $70 in two months -- just like it didn't justify a 400% price increase in five months, says Kedmey -- investors became fearful and overreacted to the negative news. For an investor holding Tesla stock, the question was going to be would the stock rebound. With Eidosearch, the trader is able to find similar occurrences of this price trend historically and to see how investors reacted in the next three months. Using the pattern of Tesla's stock one-year stock price through September 2013, Eidosearch discovered 55 similar stocks found based on comparing the price behavior of Tesla to all other consumer cyclical stocks in its sector throughout history. The average return in the next three months was 20%, according to Eidosearch.

1. Search by example using Tesla's price through Sept 2013
View Larger

1. Search by example using Tesla's price through Sept 2013

2. 55 similar stocks found based on comparing the dynamic price behavior of Tesla to all other stocks in its sector throughout history (overlay view of top 8 charts shown below)
View Larger

2. 55 similar stocks found based on comparing the dynamic price behavior of Tesla to all other stocks in its sector throughout history (overlay view of top 8 charts shown below)

Clients include major mutual fund companies, futures and currency firms and those on the equities and global macro thinkers --who are forecasting a day or a week out, and those forecasting six months or even 12 months out.

Like the other startups Eidosearch is hosted on the Amazon cloud. "We bring in data from various data vendors for our off-the-shelf solution, and some clients send it their data over night and we ingest it and make it searchable, and we can use it in conjunction with all of our data sets," says Kedmey.