Student programmers rewrite traders’ winning formulas

Listen to this article


The continuing democratisation of the market-making function is one of the most powerful and intriguing forces in finance. Knowledge of how securities were priced was once the province of tight-knit bands in the City of London, lower Manhattan and a handful of satellite centres, which collected monopoly profits by getting ahead of customers.

The gradual disintegration of those mobs is one consequence of the rise of electronic trading. I think this will accelerate rapidly in the next bear market in risk, over which liquidity will increasingly be provided by a horde of small market-makers, each with the computation and communications power of a mid-size brokerage of the past decade.

While the pieces are largely in place for this in the equity markets, the next fat-margined oligopoly to fall will be that of credit traders. For now, they are concentrated in the big banks, insulated from competition by the moat of counterparty risk. When credit trading is based on central clearing houses rather than dealer-customer trading lines, and every customer has to put up the same risk-adjusted collateral, that business will also have to be shared with the small professional trader.

At this point a professional from a big firm will patronisingly object that the laptop-armed home trader will lack the nerve, capital and skill to maintain liquidity during a meltdown. Only a blooded member of the tribe, ideally with a credit line from a central bank, will be able to do that.

That is why the results from a trading competition run by Interactive Brokers are so interesting. For eight weeks, from January 15 to March 9, the electronic brokerage ran a simulated trading competition, called the Collegiate Trading Olympiad, open to university students, each starting with $100,000 in notional capital.

Real-time price series were used, and the notional accounts were charged brokerage fees. The students had to execute at least 25 trades, which reduced the chances of making a single lucky investment, execute trades based on an automated program and do the programming themselves. The winner received a $100,000 prize, followed by two $50,000 prizes and a series of $10,000 and $1,000 prizes, with 27 winners in all.

Obviously there was something in it for IB, which was not so much to find new customers as to find new programmers. The firm earns much of its profit doing programme trading for its own account in electronic securities markets.

What makes the Olympiad results interesting was the eight-week period chosen, which included the late February correction. Were the student programmers able to “act” as marketmakers, in the sense of providing liquidity, during the correction? And did they do so in a systematic manner?

That was clearly the case for the winners. I spoke to the top two contestants, Brian Eckerly, who recently graduated in electrical engineering from Ohio State University, and Konstantinos Tsahas, a part-time student in financial engineering at Baruch College in New York. Mr Eckerly “earned” $294,190 on his $100,000, while Mr Tsahas recorded a $160,524 notional profit.

Mr Eckerly, 22, developed a strategy based on the S&P 500 index. He is not publishing his trading algorithm, so don’t ask, but he says: “It seeks to take advantage of price extremes and corrections in both up and down markets. It’s based on contrarian mean reversion.”

The natural question about any winning system is how long it will work before the environment changes and invalidates the program. Mr Eckerly is rationally modest about his algorithm. “It’s hard to say what a trading system will do, because if it is consistent then others will take advantage, so you have to adapt.”

He was more highly leveraged than he would have been with his own money but, given that it was, after all, notional capital and a short contest, that made sense. “My algorithm anticipated a decline, so I profited from the correction.”

Essentially, his tool identified the main trend in prices and which prices were just noise around it. Then he took the other side of the “noise” movements. This is exactly what a professional trader does with his “instinctive” decisions. The difference is that Mr Eckerly’s program does it automatically and, given cheap computers, broadband and automated execution, just as quickly as a big trading desk.

Mr Tsahas, also an engineer, owns a construction business in Queens, New York. He is a Greek immigrant in his 40s, from the island of Ikaria. “My strategy,” he explains, “is a pullback strategy. If you buy a market on a 2 per cent pullback, rather than waiting for a breakout to the upside, it pays better.” He uses a Bollinger band filter, which means he measures price movements in terms of one standard deviation band above or below a trend line. “Because of the liquidity in markets the last few years, it doesn’t pay to take that strategy for short sales, so I was using a long-only strategy.”

In other words, like Mr Eckerly, Mr Tsahas made money leaning against short-term trends in securities markets. That is what professional marketmakers do.

After the next bear market, much of the work on those glass-sided floors will be performed by lots of people such as Mr Eckerly and Mr Tsahas. And the key skill won’t be vague trading instincts, or the ability to worm into old-boy networks, but the rapid writing and rewriting of trading programs.

Copyright The Financial Times Limited 2017. All rights reserved. You may share using our article tools. Please don't copy articles from and redistribute by email or post to the web.