The Gateway to Algorithmic and Automated Trading

Outside looking in

Published in Automated Trader Magazine Issue 29 Q2 2013

To outsiders, the algorithmic trading industry is a citadel - big, rich, and closed to tourists. Assembling the tools needed to get behind the walls is tough, laborious and time-consuming. So what are the prospects for interlopers to prise open a gap and get inside?

Outside looking in

Quantopian, a start-up that launched in January, believes it's possible. The theory: the interlopers just need someone to provide them with the right tools. The goal is to unleash the aspiring quant traders among the cohort of scientists, mathematicians and geeks who are not working in the finance industry now. Somewhere in that crowd, there should be ideas for brilliant algorithmic trading strategies.

Open-source software is already used widely by quants, but the professional trading industry has stopped short of sharing the tools they build with that software. If all goes according to Quantopian's plan, open-source tools could be disruptive for an industry that has been responsible for its own fair share of disruption. It could change the makeup - and the terms - of the quant labour pool and infuse the community with new styles of trading.

The venture has its sceptics, but a couple of big names have bought into the theory. Quantopian's backers include Getco, one of the world's largest automated traders, and Spark Capital, a venture capital firm that looks for investment in disruptive technologies. Past investments include Tumblr and Twitter.

Andrew Parker

Andrew Parker

"What Quantopian is trying to do is take tools … that are now only in the hands of quant traders in huge institutions and democratise them,"

"What Quantopian is trying to do is take tools … that are now only in the hands of quant traders in huge institutions and democratise them," said Spark Capital principal Andrew Parker. "That's a really disruptive offering. If you look at the idea that there are 10,000 people on Wall St, they're not any cleverer than the 100,000 engineers that come out of IIT in India every decade, but those people don't have access to the tools."

(IIT is the group of Indian universities which teaches engineering.)

Former Getco quant Fred Monroe is also a fan and a backer. He introduced Quantopian to his old employer after he came across the company while looking for angel investing opportunities. For him, the approach strikes a chord because it echoes his own experience.

He was a programmer working outside the financial industry when he started to tinker with automated trading about 15 years ago. There were no tools available, so he was trying to use an online brokerage account designed for much more basic trading. "It was very painful," he remembers with a laugh.

Later, Monroe started his own prop trading firm, but still felt he was pushing against big barriers to try to compete with other people in the marketplace on a level playing field. "I always felt I was trying to get inside the ropes," he said. Quantopian's tools are something he wished he'd had when he was starting out.

The gap between the idea and the execution of an algorithmic trade is almost impossible to bridge for anyone outside the wall, according to Quantopian founder John Fawcett.

"The realisation I came to is that algorithmic investing is an algorithm production problem," he said. "You need to create a huge number of algorithms and explore a lot of them to find one that's going to work, and it takes a lot of very complicated work to get there. It's by no means easy to come up with a great idea, but what I think makes it almost impossible today is that you also have to have a world-class systems developer to build the infrastructure you need to actually try it."

Boiling out the highly profitable idea from the ocean of others that don't pan out is time-consuming. "From my point of view, Quantopian is ramping up the amount of productivity that can be applied to this problem," Fawcett said.

The Quantopian offer now is a tool for building algorithms, a back-tester the company built which they have now made open source, and 10 years of minute bar US stock data, along with a community for sharing ideas, algorithms, tools and data. Everything is free; the open-source back-tester, Zipline, always will be, Fawcett said. Users can choose to share their algorithms or keep them private, and the target is low frequency trading, not high frequency. Most algorithms tested on the company's tools have position hold times of several days.

The company is not yet generating revenue. When it does, first by offering connections to existing brokers so Quantopian users can trade, it will join a competitive field. Brokers and online services competing in the retail investor space offer a wide range of different tools, including back-testing.

NinjaTrader, which started in 2004, is one of them, offering charting, analytics and trade simulation on a free platform, along with a community of users, to a user base which is mostly day traders.

"What Quantopian is doing is a validation of what we're doing," Ninja Trader product manager Josh Peng said. "People just want the tools; they want to be able to run whatever strategy or technique they want to test on. We've always known that people want to do that without some restriction like 'I need to do that within 30 days of a free trial', or, 'I need to pay thousands of dollars for a software licence fee'. So to us it seems they're following along the same strategy."

Ninja Trader generates revenue via charges for transactions passed to brokers, sales of its data and sales of its trading platform to brokers. Some in its community also develop customised engines on Ninja Trader's existing tools and platform, which are sold or provided free to other users. The question, however, is whether something as specialised as algorithmic trading tools can work with the same business model that has thrived with the rise of retail investing.

"I tip my hat to them," said Dary McGovern, managing director of UK-based Time to Trade, a platform offering tools for retail investors. "If somebody wants to serve that fraction of a trading audience, that's great. I applaud them for doing it." But he is sceptical about the size of Quantopian's potential customer base in a crowded market.

Quantopian, though, also has a different target in mind. "The competition is not retail brokers," Monroe said, "although it would be easy to conclude that, because at this stage it might look a little bit like Trade Station or other offerings like that. But my hope and dream is that the real competition is people who make markets and manage money professionally, and the quants who work for those companies."

Fawcett said in a recent post on the Quantopian community site that the company's vision, along with generating revenue from transactions, was also to help quants connect with backers by providing back-testing results that could be compared to peers, auditable performance track records for real trading, and massive global distribution.

"Capital introduction fees today border on the predatory, because new managers have so many hurdles to be marketable to institutional investors," he said. "This is another area where I think we can create profit for Quantopian while lowering costs and eliminating barriers for quants."

The Quantopian offer now is a tool for building algorithms, a back-tester the company built which they have now made open source, and 10 years of minute bar US stock data, along with a community for sharing ideas, algorithms, tools and data.

Fawcett wants Quantopian to challenge the secrecy that now surrounds algorithmic trading systems. "The thing that we wanted to explode, which I think is a myth on Wall St, is that 100% of the system, the back tester, the simulator, the research environment, the execution environment and the algorithm need to be secret," he said. "That's utter nonsense, and completely outdated. If it ever applied, it only applies to high frequency where every detail does matter."

The valuable piece of intellectual property, he maintains, is the relatively small amount of code in the algorithm itself - 100 or 150 lines of code. "What should be shared is the back tester and the environment for the algorithm, which in our case is 30,000 or 40,000 lines of code that provide the rest of our system." In the market now, black-box algorithms are tested on black-box back-testers, making comparisons very difficult.

Quantopian flowchart

Quantopian flowchart

Dr Richard Bentley, vice-president of capital markets at Progress Software, said the Quantopian approach fits with some of the existing trends in algorithmic trading. "I've seen an awful lot lately about lack of transparency in broker algorithms - how does the buy-side know they're getting a good price, how do they know they're getting the best deal," he said. The approach also played into the trend for the buy-side to take more control from the sell-side by bringing more activity in-house.

But Bentley is sceptical about the prospects for generic tools. "Generic back-testing sounds wonderful, but one of the things we're seeing is that folks are really investing in this area, which I call validation," he said. Verification testing checks that algorithms perform as they are supposed to; validation tests the more complex problem of whether the idea behind the algorithm makes money in a highly nuanced, fragmented and dynamic market landscape. That landscape posed big challenges for development of generic tools, he said.

Quantitative derivatives trader Tim Meggs says the appeal of Quantopian is the ease of implementing ideas. Meggs was executive director of derivative trading at CIBC World Markets, and last month took on the unpaid role of Quantopian's "quant in residence" during gardening leave before starting his next job.

"People who like to tinker with things will naturally be drawn to it," he said. "You see it in the social media responses to Quantopian. When people find it they go 'Oh my goodness, this is going to take up so much of my leisure time, there goes my weekends'. I think there are a lot of people from science and mathematics backgrounds who see financial data as an interesting time series."

John Fawcett

John Fawcett

"The thing that we wanted to explode, which I think is a myth on Wall St, is that 100% of the system, the back tester, the simulator, the research environment, the execution environment and the algorithm need to be secret,"

Stephane Leroy, head of global real-time solutions at S&P IQ, raises another issue: the ability of an open source platform to be updated and maintained.

"I have seen a lot of open-source initiatives and I think it's excellent for the industry. But the open source has pros and cons. And I've seen also a lot of these open-source initiatives just dry up," Leroy said.

He said maintaining the technology can be complex. "And, you know, we are in an industry where people do not want really to share what they think is really an upside, or an additional value of what they do," he said. "At the end of the day, do I really believe in open source for model development, for data normalisation and so on?" History, he said, has suggested otherwise.

If Quantopian is disruptive, Meggs believes, it will be because its tools, especially its open-source back-testing software, lower the barriers to entry for people who have the relevant skills but are not working in an organisation with proprietary tools.

Fred Monroe, the retired former Getco quant, shares that vision of the future, but expresses it in terms that might be a bit scarier for the current algorithmic trading industry.

He grew up in Flint, Michigan, the town near Detroit where General Motors was founded but which is now more famous for its dwindling population, abandoned buildings and government in financial crisis.

"I watched a lot of friends and family think that the good times in auto would last forever, and make decisions that later came back to haunt them," he said. "I think the same thing is true on Wall St. There's a lot of people getting paid a lot of money to do something that maybe other people could do, and would be quite happy to do a lot cheaper. There are artificial barriers to keep it that way. I know I would've done my job for less money, and I'm sure there are people from other countries who would have been willing to do it for less. So I think that's where things are heading, and companies like Quantopian will probably benefit - that's my guess."

Quantopian's Fawcett says there are two reasons quants now go to work for established firms, rather than striking out on their own to exploit their personal trading strategies. Graduates choose firms because the opportunity to learn from mentors is valuable, but the other reason is that an established firm provides all the business infrastructure needed to build and use trading algorithms.

Recent heat map displaying where new users were added

Recent heat map displaying where new users were added

"They give up quite a bit in exchange for that," he said. "The intellectual property of everything they create belongs to the firm, and the individual doesn't capture all the benefit. I think what we're trying to do is to empower individuals, so that the balance of power between firms and quants can shift pretty dramatically."

Big established firms might not think that was a threat. But Fawcett said the balance of power would tilt towards quants, as the data beyond the universe of market data would become increasingly important for investment decisions. The last 20 to 30 years in algorithmic trading had mostly been about automating decisions on entering and exiting trades; in the future the kind of algorithms now coursing through time-series data would trigger investment decisions based on non-market data.

"The over-hyped big data trend - the kernel of truth that's in there is that companies in all industries are discovering they have hoards of data, and it's really quite valuable," Fawcett said. "The severe shortage right now is people who are capable of analysing it."

The disruptive power of Quantopian is all about unlocking the power of the quants. "We're on a trajectory that's really exciting to the talent, and if you understand the quantitative investing business as a talent business, clearly that's going to be very destructive."