Cover Story: Defending the Turf
Sellside Moves to Protect Buyside From HFT Onslaught
Traders Magazine Online News, May 12, 2011
Talk about the power of television.
On Monday morning, Oct. 11, 2010, the phones in Credit Suisse's Advanced
Execution Services department were ringing more than usual. On the other end were
anxious customers asking the big broker what it was doing to protect their orders from predatory black-box traders. They had all watched "60 Minutes" the night before, when the popular television program explored the world of computerized trading and the role of high-frequency traders.
Reporter Steve Kroft interviewed Joe Saluzzi, co-head of agency brokerage Themis Trading, who told Kroft that high-frequency traders were using their superior firepower-i.e., speed-to push prices up or down, to the disadvantage of the institutional investor. "They're parasites who exploit technological advantages to suck money out of the market," Saluzzi told Kroft. "They add no value." The trader added that he "spots signs of predatory behavior every day."
The following week was a busy one for AES head Dan Mathisson. "Everyone was calling after seeing that piece," Mathisson said. "We had several clients come to visit us. Many wanted information because they had to make presentations to their boards of directors."
According to buyside traders surveyed by TABB Group, the most important market structure or regulatory issue they face is high-frequency trading. Nearly half the traders interviewed for TABB's 2010 survey felt HFTs were a problem, up from 29 percent the prior year.
The biggest concern, according to the traders, was that the marketplace had become an "uneven playing field" and the speed advantage of HFTs worked to their disadvantage. HFT trading now accounts for at least half of all industry volume, according to some estimates. In TABB's study, about one quarter of the traders said the presence of HFTs had increased their trading costs.
The charge most often levied at HFTs is "gaming," or using ill-gotten information to out-trade others. (See sidebars for details.) Brokers and analysts caution, however, that the vast majority of trading done by high-frequency traders is not predatory. While stories about prop shops sniffing out institutional orders in dark pools or leveraging co-located trading engines make for juicy headlines, they comprise only a small part of HFT trading, sources say.
The more common HFT trading strategies involve market making or some form of arbitrage. These strategies are considered beneficial for the market as a whole, as they supply liquidity and keep prices close to fair value.
Still, that's cold comfort for the buyside trader trying to get a large order done. Since the market maker or arbitrageur is likely better informed about the short-term direction of the stock, the buysider trades with them at his peril.
Between the buyside and the markets, of course, is the broker. Money managers are increasingly asking their brokers about the steps they are taking to protect their orders from run-ins with HFTs. They want to know which venues their orders are touching and how they are interacting with those venues.
The sellside has not been idle. In the past year, there have been a flurry of announcements from brokers rolling out new technology and upgrading older technology. They are developing new algorithms, smart order routers, and intelligence-gathering software. On top of that, the sellside is beginning to supply their customers with the details of their order-routing methodologies and is spending more time coaching their clients on the best trading techniques.
At least two brokers are trying to turn the tables on HFTs by using their own trading tactics against them. Both Pipeline Trading and Deutsche Bank recently debuted short-term price-prediction technology that is standard issue at HFT firms.
The technology, otherwise known as short-term alpha models, is used to predict the direction the price of a stock will take in the coming seconds or sub-seconds. That gives HFTs an information advantage, meaning the buyside faces "adverse selection" risk when trading with them. In general, adverse, or negative, selection means trading with the wrong party at the wrong time.
The "wrong" party is typically one with superior information. The "wrong" time may be sooner or later, depending on the direction of the stock. "You may have traded too slowly when you should've traded more quickly," explained Henri Waelbroeck, director of research at Pipeline. "Or you traded too quickly when you should have traded slower."
The problem is especially acute when the buyside trader is posting limit orders. This passive strategy allows him to save the cost of the spread, but opens him up to competition and better-informed contra parties.
If an HFT's short-term alpha model predicts a stock is headed down, the HFT will want to quickly sell. If a buyside trader has posted a buy limit, the HFT will hit that bid. Afterwards, if the stock drops, the buyside trader has overpaid. It's because he traded too quickly.
If the model predicts the stock is headed up, the HFT will want to quickly buy. If the buyside trader is again using a passive strategy, he may find the HFT beats him to the punch. Because of their superior speed, HFTs can often get to the head of the queue faster than others. The HFT buys first. The institutional trader buys second. He has traded too slowly.
"High-frequency traders are very efficient at capturing liquidity using co-located servers," Waelbroeck said. "They essentially intercept any sell interest coming to the market and starve the institutional algorithm from accessing that particular source of flow. The result is that institutional algorithms tend to perform more slowly in those circumstances. And sometimes dramatically more slowly."
To level the playing field for the institutional trader, Pipeline recently outfitted its customers with a short-term alpha model of their own. Called Alpha Pro, the product allows buyside traders to predict the direction of the stock over their own trading horizon. That might be in hours or days. Once they do that, they can decide on their participation rate-what percentage of the volume they want to be-for the trade.
Predicting the price of the stock plays a key role in the revamp of Deutsche Bank's Stealth algorithm, as well. "Stealth begins to level the playing field between agency algorithms and modern electronic liquidity providers." said Alex Paley, Deutsche's global head of algorithmic trading product development. "These algorithms use high-frequency trading techniques similar to the ones used by liquidity providers in making trading decisions."
While posting limit orders can be risky business, hitting bids and lifting offers is not risk-free. "If you repeatedly lift the offer, you could signal the marketplace or the liquidity providers that you have a much larger order to buy," Amit Manwani, Nomura Securities' head of quantitative and analytics products in the U.S., warned the crowd at this year's TradeTech USA conference. "They can use that information to your disadvantage by widening the quote or getting ahead of you in the buying process."
RBC Capital Markets is on the case. The Canadian broker recently rolled out a routing product called Thor that sends orders to exchanges simultaneously, rather than serially, or one after another, as many routers do.
Thor is an attempt to eliminate the information leakage that institutional traders may cause when taking liquidity from multiple exchanges. The technology assumes that the quoter is an HFT with quotes strung across several exchanges. What often happens is that once the HFT gets hit or taken on one exchange, it uses its superior speed to change its markets on the other exchanges before the buyside has a chance to access those quotes. As a result, the buysider pays more or receives less when he finally does reach those destinations. But under Thor, which routes to each exchange simultaneously, that can't happen.
Kelly Reynolds, director of trading at Toronto-based Hillsdale Investment Management, is a user. "It's real-time latency adjustment for orders," Reynolds told Traders Magazine in February. "For what it does, I am very satisfied."
Despite the effort brokers are putting into combating the speculators, some on the buyside are skeptical the tools will do much good. "It's all well and good that brokers are continually refining their anti-gaming tools," said Craig Jensen, head trader at Connecticut's Armstrong Shaw Associates. "But when you're competing with technology, it won't be long before the improved logic is gameable."
The buyside trader's world would be a lot simpler, of course, if the trading venues he frequented simply banned speculators altogether. A few dark pools do this, but no lit pools do. Federal regulations prevent exchanges from barring anyone. ECNs can pick and choose who they let in, but most take all comers. That's about to change.
At press time, Credit Suisse was expected to launch a new ECN called Light Pool that would cater exclusively to institutional traders and bar high-frequency traders from the mix. It would also refrain from certain exchange practices that came under criticism for leaking information.
"Exchanges let anybody in," Dmitri Galinov, head of liquidity strategy for Credit Suisse's Advanced Execution Services group, said at TradeTech. "So it becomes very difficult to trade. That's why we created Light Pool. By using a formula, we will figure out who is toxic."
That formula is already being used to ferret out the "opportunistic" traders in Credit Suisse's dark pool, CrossFinder. By looking at the types of trades and the participants' trading horizons, Credit Suisse is able to give its money manager clients more information about who they might encounter in CrossFinder.
Light Pool will differ from existing exchanges in several ways. First and foremost, it will ban those traders deemed opportunistic. HFTs will still be able to trade against Light Pool quotes via the open market, but they will not be permitted direct access.
Light Pool will also exclude certain order types popular with HFTs-such as the "post-only" order-and certain information from its proprietary data feed. That feed will transmit quotes, but not trades, depriving hyper-fast speculators of valuable information. Both Nasdaq and BATS exchanges came under criticism last year for transmitting sensitive information in their proprietary data feeds.
Although most of their liquidity is supplied by high-frequency traders, the public markets don't have a monopoly on gaming and adverse selection. Dark pools are swarming with HFTs, as well. "Much of the market-making flow is making its way into these internalization engines," Jatin Suryawanshi, head of global quant strategy at Jefferies & Co., told the crowd at TradeTech. "So essentially, you have a quasi exchange within your ATS."
Suryawanshi explained that relatively few trades are done at the midpoint of the bid and offer in dark pools. About 70 percent are actually done at prices within 5 percent of the bid when selling and offer when buying. These order types are known as "pegging with an offset," where the execution price is pegged to the NBBO, for instance, offset by a tiny amount-maybe 1 mil or 5 mils-of price improvement.
"If you post passive liquidity in the dark pools and you want to capture the spread, the market makers will front-run you as they give you a meager hundredth of a penny price improvement and to get the first look," Suryawanshi explained.
There are ways to beat this game. Dark pool operators have long allowed buyside traders to set minimum fill rates as a means to avoid trading with potential predators. Because most HFT pinging is done in lots of 100 shares or so, the pools let traders set a floor under the size of any order they interact with. Algorithms, sometimes known as dark aggregators, typically take advantage of these settings.
At least one broker is taking the minimum fill rate defense one step further. Weeden & Co., in partnership with Pragma Securities, which built the first dark aggregator in 2007, recently updated their OnePipe tool to set minimum fill rates dynamically.
With the recent release of the third version of OnePipe, Pragma has made the detection of liquidity in the pools more science than art, as it was previously. Using real-time monitoring of a pool's liquidity conditions, OnePipe can better estimate just how much liquidity is available. That means Pragma can now pinpoint the appropriate pool-out of a possible 30 or 40-with greater certainty of a fill.
At the same time, the new information allows Pragma to more precisely adjust an order's minimum fill rates on the fly. Minimum fill rates are a boon for anti-gaming, but they can result in missed opportunities if set too high. With the enhancements, if a significant amount of liquidity entering a given pool is in large chunks, for instance, OnePipe will ratchet up the minimum quantity. That allows it to avoid gaming by small orders. When incoming flow is in smaller chunks, the fill rate can be dialed down.
Dark aggregators have come under criticism in recent years. In 2009, agency brokers Pipeline Trading and Investment Technology Group analyzed the algorithms and found them wanting. They wrote papers claiming the algorithms forced institutional traders into unwanted trades with HFTs.
At least one buyside trader has been burned. "All of sudden you will put something in a dark aggregator and receive an execution at the midpoint," Ed Doyle, senior equity trader at Pioneer Investments, said at TradeTech. "You thought it was great that you didn't have any market impact. Then you move on. Other people start to play catch-up. You start to realize you had more impact from that one order than anywhere else. You tipped your hand."
Dark pool operators are not deaf to these complaints. Three years ago, at least two of the bulge brokers, Credit Suisse and Goldman Sachs, began monitoring the activity in their dark pools.
Credit Suisse, whose CrossFinder is the industry's largest dark pool, began categorizing its user base into seven groups. It let traders choose which groups they wanted to trade with. Most of the firm's buyside customers opted not to trade with HFT and other prop types, Mathisson said.
Credit Suisse recently refined its classification system, dividing the universe of its dark pool into three inhabitants: contributors, neutral parties and opportunistic traders. Contributors are mainly retail brokers sending in highly desirable "uninformed" order flow. Neutral traders are primarily institutions and constitute the vast majority of players. Opportunistic traders are the high-frequency types.
Under the new rules of engagement, contributors may trade with whomever they wish. Neutral parties will be able to trade with other neutrals and opportunistics. They will also be able to trade with contributors, if those players permit. Finally, opportunistics will be able to trade among themselves. They may also be allowed to trade with neutrals and contributors, if those players permit.
But again, most buyside traders have opted not to trade with opportunistic traders. That has resulted in very little HFT activity in CrossFinder, Mathisson said. "We operate a tighter environment than the exchanges," he said, "where you have to trade with absolutely anybody who wants to be on the other side. Exchanges have no restrictions."
Despite the recent push by brokers to safeguard their customers' orders, some execs question whether brokers are really working in the best interests of money managers.
Morgan Stanley, for example, has been critical of certain order-handling practices, arguing that many of them are either irresponsible or benefit the broker to the detriment of the client. Either way, they play into the hands of some short-term players. "The goals aren't always 100 percent aligned between the broker and the customer," said Andy Silverman, global co-head of electronic trading at Morgan Stanley.
The exec noted that many brokers indiscriminately connect to as many destinations as possible under the philosophy that all liquidity is good liquidity. En route to the lit market, they might ping as many dark venues as they can. That leaves a trail the HFTs can follow. It also might cut the broker's trading costs, as most dark pools charge less than the exchanges. "All liquidity is not necessarily good liquidity," Silverman said. "What happens when you go to 40 venues? What do you leave behind? You leave a footprint."
Butting heads with market speculators is nothing new. Tape readers, for instance, have been around for hundreds of years. Many even compare today's HFTs with the Nasdaq market makers and New York Stock Exchange specialists of yesteryear. What's different today, some contend, is the sheer number of speculators in the stock market. At half the volume, they are a force to be reckoned with.
"We've always had anti-gaming in place," Silverman said. "But as the players develop new strategies, we have to adjust to those strategies."
For more information on related topics, visit the following channels: