In 2014, the Stevens Institute of Technology set out to create a new financial market simulator to research market microstructure, starting with equity markets. Last year, the college ran a pilot contest among students using the platform, called Shift. Then at the beginning of this year, consultancy Capco joined the initiative to help further build out the market simulator’s capabilities. With the kinks worked out, they are now looking to commercialize the product.

Shift, an artificial exchange, started as an academic exercise for students of Stevens to test theories and stress portfolios in a sandbox environment, but it developed into something the school believes could appeal to the wider trading community, says Stevens professor Dr. George Calhoun.

“What helped us cross the line, even before Capco got involved and our thinking about this as a commercial product, was the reaction of firms in the industry when we would tell them about this research project,” he says. “They said, ‘Have you ever considered making this into a commercial product that you could make available?’ When we heard that a half-dozen times from people [at] Goldman Sachs and UBS and so on, we started to think that maybe there’s something there that we should be paying attention to.”

Calhoun believes that Shift—which stands for Stevens HIgh Frequency Trading Market Simulation System—can prove valuable when it comes to better examining high-frequency trading (HFT) market structure, and understanding the causes of so-called flash crashes. It could also be a useful tool for regulators and trading firms to comply with certain requirements stemming from Regulation Automated Trading (Reg AT) [see Box below].

“Back-testing is historically misleading. I think one of the reasons for that is that the market mutates—it changes all the time, and it mutates in response to new players coming in and out of the market. The Shift project is based on the idea that we’re not trying to prepackage anything,” Calhoun says. “We are trying to make this as real a simulation of the market, in all its detail and all its speed, as possible. That is an objective that I don’t think has been the motivation for some of the other products that are out there, which are more about giving a student the ability to play with a portfolio and see what happens.”

There is still work to be done, however, before Shift—which is cloud-based and written in C++—is ready for wide-scale market adoption. Leonard Langsdorf, CTO of Capco’s Digital Labs, says they still need to build a better user interface, and they are still considering a pricing model for commercialization. Stevens students have also run contests that have helped to “find all the holes,” Langsdorf says, before they commercially launch the product.

The Proving Grounds

For research purposes, Shift works with live, real-time, order-level market data provided by Refinitiv. (For when the platform expands to other asset classes beyond equities, Stevens already has a partnership with CME Group, which has provided funding for the project.) For the commercial piece of the simulator, the system replays recorded datasets of quote data. All communications on the platform—which includes a datafeed engine, matching engine, and a brokerage center—are done using the FIX protocol. Users can access Shift via a web interface or APIs written in C++ and Python.

Several artificial markets already exist—the Santa Fe Artificial Stock Market and the Genoa Artificial Stock Market are two notable examples—but Calhoun believes that Shift is significantly different from other platforms.

First, it uses a real pricing mechanism, and draws pricing from “all the major exchanges,” to provide a true understanding of market microstructure, Calhoun says.

Users can setup their own brokerages, where, for example, one brokerage with $1 trillion under management can be created and another with$1 billion. From there, the user can see how the two entities would interact, Langsdorf says.

Shift is a distributed asynchronous system. In a paper published earlier this year by professors at Stevens, it stated that in its system, “agents perform actions whenever they want to, and the central unit is constantly listening for incoming messages, with no control over when they are sent and by whom.” In this way, it better mimics a real market because in a “turn-based system,” it’s impossible for orders from low-frequency traders to arrive before an HFT order, but in a real system, sometimes “low-frequency orders operating on outdated information” do arrive earlier.

“The problem with many, many academic papers on this type of thing is that they use daily data … but then the market microstructure comes into play and you’re not going to [truly understand how a portfolio will perform over the long term] until you get into the high-frequency side [of the market],” Calhoun tells WatersTechnology.

Asset Expansion

While the current iteration of Shift only covers equities, it was designed to be a multi-asset market, as “allowing agents to trade multiple assets can potentially recreate the highly correlated markets we are experiencing today,” the paper states, pointing to the Flash Crash in May 2010 as a possible research experiment.

“Getting a good understanding of what is causing these events is a major challenge in the market today for all the participants, exchange operators, and regulators,” Calhoun adds. “Is it a fat-finger thing, an interaction between a small number of players, or a larger wave rolling through the market? I would say that the [Flash Crash] question is still an open one and there will be different answers to that for different types of technology risk phenomena that we’re going to see, but that’s the type of question that you can really look at with this type of tool.”

He adds that they have shown Shift “in an informal way” to representatives from the Commodity Futures Trading Commission (CFTC) and the Securities and Exchange Commission (SEC). “There’s a lot more that has to be done before we have definitive answers, but that’s part of the value proposition here,” Calhoun says.

Capco’s Langsdorf (pictured, right) says equities has been “nailed down,” so they will next look at futures, options, and currency, with the latter likely to be the next asset addressed. However, there isn’t an official timetable in place for that roll-out.

“We’ve generalized it enough to the instrument level now that we can start adding new instruments, such as credit swaps if we can get fixed-income data, or we can just do straight-up options and mimic short-selling concepts,” he says. “We do have the ability; it’s really just taking the time and plugging it in.”

Langsdorf says the platform is set up to ingest any data source. While Stevens has partnerships with Refinitiv and CME Group, Calhoun says they hope to add on additional exchange partnerships in the future.

“We have the relationship with CME,” Calhoun says. “We have not dialed it up yet to bring in [data for other asset classes], but … that’s something that when we’re ready to do it, it’s going to be there for us to bring in. We also have [a relationship] with Nasdaq; I guess we don’t have one at the moment with ICE, but we haven’t really attempted to go out and create those channels quite yet.”

The key to further expansion will be landing that first investment bank, proprietary trading shop, asset manager, and/or hedge fund, to both give the project industry validation and to work out the user interface.

“It’s during that first install that you get the most information, and then it becomes a very viable product,” Langsdorf says.

### Reg AT

In late 2015, the CFTC first introduced Reg AT. While parts of the rule were applauded, its source-code provision, which requires that each market participant keep a source code repository for algorithms and make it available for inspection to any representative of the Commission or the US Department of Justice for any reason, has faced a serious backlash. That piece of the rule has been on life support, but the regulator still wants to better understand how high-speed automated traded affects the market.

The research paper written by professors from Stevens specifically pointed to Reg AT as being an area where Shift could help the regulators, pointing to the piece of the regulation that says that trading algorithms need to be tested in “laboratory conditions” before they can enter the market.

The paper states: “We believe that Shift may create an environment where algorithms can be tested and stressed in laboratory conditions. The environment may be setup so that proprietary source code may be tested adequately in absence and without participation of other market participants.”

Calhoun compares it to how the Food & Drug Administration (FDA) conducts clinical trials on vaccines before they’re allowed to legally be sold in the United States.

“You’re going to trial the vaccine in multiple phases before you put it out in the general public,” he says. “I think Reg AT has stated the need; I don’t know that we’ve yet had the ability or opportunity to make the formal case that we think we’ve got this capability … but that is a step that I think will come up very soon in our pathway here.”

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.waterstechnology.com/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

#### Waters Wavelength

###### Waters Wavelength Podcast: Episode 236 (Market data, ESG, and interoperability)

Wei-Shen and Tony talk about WatersTech’s three biggest stories of the week.