High-Frequency Trading Is Newest Battleground in Crypto Exchange Race (#GotBitcoin?)
High-frequency trading (HFT), a longtime and controversial practice in traditional markets, is becoming commonplace in crypto, too. High-Frequency Trading Is Newest Battleground in Crypto Exchange Race (#GotBitcoin?)
Placing trading servers physically close to exchanges’ matching engines can win an edge on speed. This helps HFT firms make large profits in the legacy markets.
Crypto exchanges such as ErisX, Huobi and Gemini are trying to attract large algorithmic traders with colocation offers.
Demand for the service is high, but its benefits are a matter of debate, due to the structure of the crypto market.
A handful of cryptocurrency exchanges are rolling out the red carpet for high-frequency traders.
Huobi, based in Singapore, and ErisX, in Chicago, have separately begun offering colocation, in which a client’s server is placed in the same facility or cloud as the exchange’s, officials at each exchange told CoinDesk. This allows those investors to execute trades up to a hundred times faster, giving them an edge over the rest of the market.
These exchanges join Gemini, which was one of the first crypto firms to offer colocation at a popular data center in the New York area, and is about to expand the option to include a second site in Chicago.
Notably, none of these exchanges charges for the service, seeing it as a way to differentiate themselves. “It’s our competitive advantage,” said Andrey Grachev, head of Huobi Russia, the exchange’s Moscow client office.
To be sure, such accommodations remain rare in crypto, which historically was dominated by individual traders and only recently began to draw interest from institutional investors such as hedge funds and family offices.
But the exchanges’ moves are a sign that high-frequency trading (HFT), a longtime and controversial practice in traditional financial markets, is slowly entering the crypto sphere. And though “bots” have been present in crypto since the days of Mt. Gox, colocation takes algorithmic trading to a different level.
Eric Wall, former crypto and blockchain lead at Cinnober, a financial technology company acquired by Nasdaq, told CoinDesk:
“It’s big business, everyone I’ve been speaking to that runs an exchange mentioned being approached by Wall Street types with these kinds of requests.”
Most crypto exchanges are not ready to satisfy this demand, Wall said. These are “very new concepts to many retail-focused exchanges with no experience of the traditional world, it seems.”
800K Trades A Day
In the six months since Huobi opened its Russia office, around 50 clients have taken advantage of its colocation service by locating their servers in the same cloud and using the same domain name service (DNS) as the exchange, according to Grachev.
The option allows these clients to make trades 70 to 100 times faster than other users, he said. “One of our clients makes about 800,000 trades a day, and there are more and more such clients.”
Unlike many crypto exchanges that use cloud-based servers, ErisX has a hardware matching engine, located in the Equinix data center in Secaucus, New Jersey, said Matthew Trudeau, the exchange’s chief strategy officer.
The same facility houses the matching engines of a range of major traditional exchanges, brokers and trading firms, Trudeau told CoinDesk, so traders that colocated servers in the data center can connect to ErisX’s matching engine there. (The firm launched spot trading in several cryptocurrencies in April and recently obtained regulatory approval for futures.)
Gemini, founded in 2014 by Cameron and Tyler Winklevoss, also houses its primary trading platform at Equinix and offers colocation there. The exchange plans to offer another colocation option soon in Equinix’s Chicago data center, where multiple stock exchanges — and their HFT customers — keep their hardware, according to Gemini’s website.
In a statement, Gemini’s managing director of operations Jeanine Hightower-Sellitto said the exchange “offers a variety of connectivity options to suit our customers’ needs. Each option is available to all of our customers free of charge.”
Coinbase, the leading U.S. crypto exchange, almost entered the fray, but this year closed down its Chicago division that had been working on services for high-frequency traders, including colocation. At the time, the exchange cited its prioritization of other institutional services.
The company declined to comment for this article. (Gemini, which just opened a Chicago office, hired some of Coinbase’s former employees there.)
All of this invites the question of whether HFT, given its history on Wall Street, could exacerbate problems in the opaque and volatile crypto markets.
As depicted in Michael Lewis’s book Flash Boys, algorithmic stock traders placed their servers in the physical vicinity of exchanges’ to execute trades faster than other investors and make profits on arbitrage between markets in fractions of a second.
The issue with HFT, as explained by Lewis, is that in a market where some players can perform trades hundreds of times faster than ordinary users, they get an unfair advantage and leave ordinary, non-algorithmic traders with inferior price options.
Another problem with HFT, according to a 2011 report by the International Organization of Securities Commissions (IOSCO), is that it can dramatically increase volatility in markets.
In particular, it contributed to the so-called Flash Crash on May 6, 2010, when the prices of many U.S. securities fell and recovered dramatically in minutes, exposing ordinary traders to a higher risk which they couldn’t manage as quickly as HFTs.
High-speed trading has led to other technical glitches that cost companies hundreds of millions of dollars, the Federal Reserve Bank of Chicago wrote in 2012, noting that “some high-speed trading firms have equity ownership stakes in certain exchanges.”
However, ErisX’s Trudeau (who, it should be noted, was one of the early employees of stock exchange IEX, the heroes of Flash Boys) argued that high-frequency arbitrage and automated trading, in general, can benefit markets.
They are helping to narrow the price spread between different exchanges over time and make markets more efficient – including the crypto market, Trudeau said, explaining:
“This phenomenon has occurred in other asset classes as trading has become more electronic and more automated. Market makers and arbitrageurs are able to trade more efficiently, which improves price formation, price discovery and liquidity. Arbitrage opportunities may become fewer and more fleeting, which is a sign of a more efficient and maturing market.”
It’s important, however, to check if the exchanges and high-frequency traders strike deals with preferential terms which are not disclosed to the market, he noted.
As for ErisX, it “offers transparent, standardized pricing and connectivity options for our customers. All customs are offered the same terms of access and fees,” Trudeau said.
For its part, Huobi tries to make sure all users “compete on a level playing field,” said the exchange’s head of global sales and institutional business, Lester Li.
Li Told CoinDesk:
“Our users know that we monitor for any abusive trading activity. We also continually remind users that there will always be risks when you trade, that is why we strongly recommend users to trade within their means and be mindful of the risks involved.”
Still, other exchanges contacted by CoinDesk made a point of saying they don’t do anything special for algo traders.
A smaller exchange tailored for institutional clients, LGO Markets, which launched earlier this year, took the opposite approach, deliberately slowing the trading process for everyone, according to CEO Hugo Renaudin.
Before getting matched, the orders are gathered into batches and the hash of every batch gets recorded in the bitcoin blockchain — each batch takes around 500 milliseconds to form, so this serves as a “speed bump” for trades, Renaudin said. As a result, “every trader has the same feedback on the activity of the platform.”
Taking a similar stance, Kraken’s vice president of engineering, Steve Hunt, told CoinDesk the exchange doesn’t do anything differently for HFT customers.
“We want all customers regardless of size or scale to have equal access to our marketplace,” Hunt said.
Binance, the world’s largest crypto exchange, is not considering offering colocation, account manager Anatoly Kondyakov told attendees of a recent “elite investor” meetup in Moscow. He gave two reasons.
First, “we’re trying to protect retail customers,” Kondyakov said, answering a question from the audience. Second, colocation means an official presence in a particular jurisdiction, he said, which Binance is not willing to do at the moment. (Binance is known for its deft regulatory arbitrage.)
Still, others said the crypto market hasn’t caught up with the traditional financial world to the point where offering colocation services to HFT firms would make much sense.
“Currently, the crypto market structure is still developing. HFT, in the context of equity and FX markets, does not really exist,” said Wilfred Daye, head of financial markets at San Francisco-based exchange OKCoin.
Traders coming into crypto from the traditional markets do ask for colocation, he said, but “the ask is one-off, not a popular ask in crypto,” so OKCoin doesn’t offer this service.
David Weisberger, сo-founder and CEO of market data platform Coinroutes, has another reason to be skeptical about HFT in crypto: this market is so much more dispersed and volatile that what works with stocks just won’t with bitcoin.
The concept of HFT front-running is irrelevant in crypto, Weisberger said, where the prices vary between different exchanges much more than in traditional markets:
“In futures or equities, with relatively large minimum quote variations, the bid offer spread is often stable with a lot of bids and offers at the same price. In that circumstance the fastest gets to be at the front of the queue whenever the price changes. Those orders at the front of the queue are profitable, while the ones at the back are not. In crypto, the tick size (price variation) is so small, it is easy to be ‘first’ by paying a slightly higher amount, so no need for incredible speed.”
Plus, crypto exchanges are so scattered around the world that there is no point in “being colocated to one exchange and still having to wait seconds for Binance to update,” Weisberger added.
The reason there is demand for colocation at crypto exchanges, he concluded, is simply human nature:
“People always fight the last war. People do what they are used to.”
Ultrafast Trading Costs Stock Investors Nearly $5 Billion A Year, Study Says
U.K. regulator’s study says ‘latency arbitrage’ imposes a small but significant tax on investors.
High-frequency traders earn nearly $5 billion on global stock markets a year by trading shares at slightly out-of-date prices, imposing a small but significant tax on investors, a new study says.
The study—released Monday by the U.K.’s financial regulator, the Financial Conduct Authority—sheds light on a controversial practice called “latency arbitrage,” in which ultrafast traders seek to react to fresh, market-moving information more quickly than others can.
Such information could range from corporate news to economic data to price fluctuations in other stocks or markets. Electronic trading firms invest in sophisticated technology, such as networks of microwave antennas linking exchanges thousands of miles apart, to process such information and execute trades in millionths of a second.
The FCA’s study comes as politicians in both Europe and the U.S., including Sen. Bernie Sanders (I., Vt.) and Sen. Elizabeth Warren (D., Mass.), have pushed for a financial-transaction tax, a policy aimed in part at curbing high-speed trading. The study could also fuel efforts by exchanges to restructure their markets to limit latency arbitrage—for instance, by introducing split-second delays before trades, known as speed bumps.
Many experts say latency arbitrage raises costs for investors by making everyone in the markets less likely to post competitive price quotes for stocks, knowing that those quotes could get picked off by speedy traders. That, in turn, means investors get slightly worse prices whenever they buy or sell shares, traders say.
Advocates of high-frequency trading say concerns about latency arbitrage are overblown, and they have disputed past studies that suggested the practice reaps billions of dollars a year in profits. Precise data on latency-arbitrage profits is unavailable because of the opaque nature of most high-frequency trading firms.
The FCA’s study called latency arbitrage a “tax” amounting to 0.0042% of daily stock-trading volume. The study’s authors derived that figure by examining about two months of activity at the London Stock Exchange in 2015, using a type of raw trading data that hasn’t been previously studied by researchers.
Although that’s a tiny number—less than one-half of one-hundredth of 1%—the study’s authors say it adds up. If latency arbitragers made a similar rate of profits elsewhere, they would have earned $4.8 billion on stock exchanges around the world in 2018, including $2.8 billion at U.S. exchanges, the FCA study found.
Moreover, the impact of latency arbitrage is unevenly distributed, with big investors facing the greatest costs, according to the study’s authors, who include two FCA researchers and a professor at the University of Chicago’s Booth School of Business.
“The latency arbitrage tax does seem small enough that ordinary households need not worry about it in the context of their retirement and savings decisions,” they wrote. “Yet at the same time, flawed market design significantly increases the trading costs of large investors, and generates billions of dollars a year in profits for a small number of HFT firms and other parties in the speed race,” they added.
The study found that about one-fifth of trading activity at the LSE was concentrated in brief “races” between firms seeking to engage in latency arbitrage. In such races, two or more firms attempt to trade the same stock at the same time, but only the first can profit by being the quickest to execute its trade.
During the period examined in the study—43 trading days from August to October 2015—about 22% of trading volume in FTSE 100 stocks took place in such races, which on average lasted 81 millionths of a second, the study found. The FTSE 100 is the U.K.’s large-cap index, with such companies as BP PLC and Vodafone Group PLC.
The FCA’s study relied on more than 2 billion electronic messages that trading firms sent to the LSE, or that the exchange sent to traders, during that period.
Such data—which hasn’t been used in past studies of latency arbitrage—showed failed attempts to trade as well as actual trades. That allowed the authors to reconstruct the tiny bursts of activity in which multiple firms raced to seize the same brief profit opportunity.
The data also showed only a few firms can profit from latency arbitrage, a finding that likely reflects the cost of building and maintaining the technology needed for ultrafast trading.
More than 80% of races in FTSE 100 stocks were won by the same half-dozen firms, the study found. It didn’t identify the firms in question.
High-Frequency Trading Is Newest, High-Frequency Trading Is Newest, High-Frequency Trading Is Newest, High-Frequency Trading Is Newest, High-Frequency Trading Is Newest, High-Frequency Trading Is Newest