Student's research could shake up Wall Street

Mar 12, 2013 by Kathy Hovis
Student's research could shake up Wall Street
Aleksander Zvorinji '13 is creating an algorithm that could outperform many financial analysts in predicting the stock market. Credit: Brian Beaudry

Financial analysts, take note! Aleksander Zvorinji's senior thesis may someday make them break out in a sweat.

Zvorinji '13, a College Scholar majoring in economics, psychology and information engineering, is working to complete an algorithm that could outperform the markets with a high degree of confidence.

"If this works, it could have big repercussions," Zvorinji says with a smile. "It could tell people the top stocks and how much to invest in each."

The idea could be stress-inducing for , whose jobs might be in jeopardy if the algorithm works, but, ironically, Zvorinji needs them to continue their work for his algorithm to succeed. It's based on aggregating the predictions of more than 5,000 North American analysts.

A native of Sweden, Zvorinji chose to attend Cornell, wanting a break from the rigid educational structure of Swedish schools and the chance to combine his interest in math and science into a career with real-world applications, which he found in economics.

Once here, he discovered the College Scholar Program and jumped at the chance to work on a project combining his interests in economics, , math and psychology. "I knew this would allow me to discover what I really want to do," he said.

David Easley, the Henry Scarborough Professor of Social Sciences in the Departments of Economics and Information Science, and Jon Kleinberg, the Tisch University Professor in the Department of Computer Science and chair of the Department of —two of Zvorinji's advisers—were impressed with his research idea, but knew it would be a massive undertaking with no guarantee of success.

"As with most empirical projects of this sort, Aleks won't know until he's nearly done whether it will work," Easley said. "But he is a good example of students who are blending several interests—in Aleks' case computer science and economics—and using theory and statistical tools to analyze real-world problems."

The economics department is working to expand its opportunities for undergrads to tackle research projects like this one, Easley said, a challenge in a department that graduates 200 majors each year.

Zvorinji's initial proposal focused on the idea of collective intelligence: Given a large enough group of people, taking the average of their solutions to a problem will lead to the exact answer. Zvorinji tweaked the idea a bit to tackle one of most puzzling problems he could think of—predicting the stock market.

"Say one analyst for Apple stock is always about 10 percent above the mark, while another is usually 3 percent below," he explained. "My work creates a 'fund' for every analyst, and then evaluates each 'fund' on its performance using an algorithm similar to Google's. At the end of the day, it is a 'fund of funds problem' approached in a very novel way."

Zvorinji is working toward a graph that allows more accurate analysts and successful stocks to rise to the top. He's hoping to finish up the model before graduation and test it using data from the past. The final product, which he knows will probably come after graduation, could be a publicly available index or a product to sell to investment firms.

Zvorinji's other thesis project involves a psychological test for color malleability: whether people's perception of color can be changed through a process that makes one color more positive than another.

Explore further: Rapid UK population growth undermines living standards, but may be necessary for economic growth

add to favorites email to friend print save as pdf

Related Stories

Metric predicts stock market 'flash' crashes

Dec 03, 2010

The May 6, 2010, stock market crash briefly erased almost $1 trillion in value and plunged the Dow Jones Industrial Average into its biggest intraday fall ever. The market recovered most of its losses within the hour, but ...

Computer programs that think like humans

Feb 13, 2012

Intelligence – what does it really mean? In the 1800s, it meant that you were good at memorising things, and today intelligence is measured through IQ tests where the average score for humans is 100. ...

Google backs research to improve social networks

Feb 22, 2011

Cornell researchers will collaborate with Google experts to improve group interaction in online social networks. The work will be supported by a grant of $800,000 from Google Inc.

Recommended for you

When rulers can't understand the ruled

11 hours ago

Johns Hopkins University political scientists wanted to know if America's unelected officials have enough in common with the people they govern to understand them.

When casualties increased, war coverage became more negative

15 hours ago

As the number of U.S. casualties rose in Afghanistan, reporters filed more stories about the conflict and those articles grew increasingly negative about both the war effort and the military, according to a Penn State researcher. ...

User comments : 6

Adjust slider to filter visible comments by rank

Display comments: newest first

StarGazer2011
3.9 / 5 (8) Mar 12, 2013
Nice idea but it will fail due to a false assumption. The poor kid has assumed 'financial analysts' predictions are better than random, they are not. GIGO
VendicarE
4.2 / 5 (6) Mar 12, 2013
The stock market is self randomizing.

Prediction is not possible if the method of prediction is generally known.

The best method of prediction of the stock market is insider trading, which of course is rampant in the corrupt stock trading industry.

baudrunner
3.7 / 5 (3) Mar 12, 2013
The only way of predicting the stock market is to trade only stocks whose values are affected by the season, as in futures, for example. Whiskey and wine sales increase during the holiday season, and Seagram stock goes up. How can a commercial stock analysis product correctly predict the introduction of a new product that increases the value of a company's stock? Is it only analyzing trading activity and making decisions based on their successes and failures? So many real time factors that have nothing to do with trading activity are built into the market that there will never be a successful software product to do the trading for you while you sit back and count the money coming in. That's just impossible.
alfie_null
not rated yet Mar 13, 2013
Sort of like a second order indexing?
Over the short term, he could end up making a lot of money, selling his analyses if nothing else.
antialias_physorg
2 / 5 (4) Mar 13, 2013
The problem is more that analysts try to outguess other analysts. So the ones that are best at outguessing others (making money off off their 'false' choices) win.

But if you take away the bad analysts then the good ones won't have anyone to make money off of.

Remember: stocks are a zero sum game (since they don't create any wealth/products). For someone to win at the stock market someone else MUST lose.

That is why the individual/small investor can only lose - because the high-power/high volume investors have the faster access and reaction times to stock information. It's like a lottery or playing roulette.
You can get lucky as a small time investor. But the longer you play the worse your chances (and you don't even start out with a 50/50 shot at making any money due to inflation).
barakn
not rated yet Mar 25, 2013
Many companies would never have gotten of the ground without using stock as a funding source, and companies are involved in the act of creation of wealth/products. Furthermore, by offering dividends, companies share profits with shareholders. So to claim it's a zero sum game is to demonstrate one's ignorance.