(Traders Magazine, August 2005) — Last month's announcement by Lehman Brothers that it had struck a deal to integrate a pre-trade analytics package into one of the leading buyside order management systems was not the best of news for Credit Suisse First Boston. The unit of the large Swiss bank-which single-handedly developed the mass market for algorithmic trading and whose name is synonymous with algorithms-is at least a few months behind its bulge-bracket competitor at the pre-trade front The lag underscores the dilemma in which CSFB finds itself.
Despite its status as the king of algorithms, CSFB is now under considerable pressure to stay ahead of the pack in offering electronic trading services to money managers.
"No one is going to give you an order because you were first," Dan Mathisson, co-head of CSFB's Advanced Execution Services (AES), says. "There is enormous pressure to innovate."
In 2001, CSFB was the first bulge-bracket player to aggressively push algorithmic trading as an antidote to the difficulties of trading stocks following decimalization. The firm, a top 10 trading house based on traded shares reported to AutEx Blockdata, dominated the landscape for the next two years.
Everything is different now. Every one of the top 10 or 15 investment banks is furiously churning out computerized trading tactics and analytical products, hoping to hold onto the order flow of their increasingly sophisticated clients.
For Lehman's top electronic trading executive, Jeff Wecker, getting content such as pre-trade analytics onto the desktops of the buyside, "is what will separate the leaders from the followers" in electronic trading.
Those are certainly fighting words.Yet CSFB is undaunted.
"Being first has given us enormous advantages," Mathisson said. "We have the most mature and reliable system. We've hit all the problems you can hit over the past four years. We've had time to refine our procedures, our processes and our technology. We've been able to improve our trading logic."
To that end, AES has been able to reduce both the number of milliseconds its systems take to execute an order as well as the average number of shares in an individual order. Routing time has decreased from 180 milliseconds to between 12 and 20 milliseconds. The average order size is down to 230 shares from 300 previously. "Smarter, faster, cheaper," Manny Santayana, co-head of AES, says. "That's the goal."
Improving its existing programs is one of three goals in CSFB's drive to maintain its edge. The others include devising new products and customizing algorithms for clients.
"Customization is becoming a significant percentage of our business," Mathisson said. "Clients know how they want the algorithm to behave." To that end, a CSFB crew will call on a customer to determine his trading goals and then propose a solution. The CSFB team includes a sales person, a trader, a quantitative analyst and a programmer.
For the record, CSFB's core algorithm menu consists of five different tactics. It has programs that achieve VWAP and TWAP; one for minimizing implementation shortfall; another for matching the price of the stock at the time the order was entered; and one for beating the closing price.
The big broker has three new products in various stages of completion: the aforementioned pre-trade cost analysis package; an algorithm for trading illiquid names; and portfolio-trading algorithms. (See sidebars.)
Mathisson and Santayana could easily rest on their laurels. In the past four years, the two have played major roles in propelling CSFB from the disgrace of an IPO trading scandal into one of the Street's top three algorithmic trading houses.
In 2002, CSFB paid out $100 million to settle claims it allocated IPO shares to its clients in exchange for unduly high commissions. Also, that year, two of CSFB's top equities executives were suspended and fined hundreds of thousands of dollars by the NASD and CSFB itself for their roles in the IPO debacle. (They have since returned to their posts.)
Since that low point, the image of the equities department has been transformed from an out-of-control Wild West show into a state-of-the-art electronic trading house. CSFB's equities group now looks more like the quant-driven Investment Technology Group than it does a typical investment bank's trading department.
AES algorithms handle, on average, half of CSFB's 200 million shares traded daily in U.S. markets, according to AES execs. That's both single stock as well as portfolio trades. However, the AES group in New York totals only 25 employees-a small percentage of the hundreds of traders and sales people packed onto CSFB's cavernous trading floor.
Along with Goldman Sachs and Morgan Stanley, CSFB dominates algorithmic trading. Each bank, according to a study conducted last year by TowerGroup, controls approximately 16 percent of the shares traded via algorithms.
Another study, released this year by the Tabb Group, which surveyed 53 buyside traders, credited CSFB with a 23 percent share. Goldman, Morgan and Banc of America Securities followed with 10 percent each. The numbers were calculated based on individual buyside traders' responses.
Even if CSFB isn't the No. 1 player by share volume-CSFB execs maintain they are-it has certainly captured the "mindshare" of traders. A recent Greenwich Associates poll of the buyside found CSFB was first among bulge brackets with electronic trading. CSFB was the only big broker to be grouped in the upper echelons of electronic trading with such specialty shops as Instinet, ITG, Bloomberg TradeBook and Liquidnet.
AES' success hasn't been lost on CSFB's senior management. At the company's "investor day" event last December, chief executive Brady Dougan singled out the equities division's leading position in "electronic delivery" as a force for growth. Electronic delivery takes in not only AES, but program trading, client analytics and transition management as well.
Dougan's plan is to segment CSFB's equities customers into three groups, depending on their relative contributions to the firm's bottom line. The "execution oriented category," according to Dougan, makes up 41 percent of the firm's client roster." These customers will get access to CSFB's computers, but not to its research or its sales traders. They contribute a whopping 58 percent of CSFB's pre-tax profits.
Going forward, only 22 percent of CSFB's institutional customers will be eligible to receive such high-touch services as research and sales trading. The rest get a mix of services.
The restructuring of the equities group -as seen by the client segmentation plan and trader layoffs of recent years-speaks to that department's sluggish financial picture. Revenues have remained flat at about $1 billion for each of the past three years.
This year started off worse than last year. In the first quarter, equities grossed $240 million, down from $347 million in the same period last year, according to CSFB filings with the Securities and Exchange Commission. Besides cash equities, the equities group also takes in derivatives, convertible bonds and prop trading.
CSFB's traditional manual stock trading operations are in decline, according to a recent 10-Q filing, because of "commission compression." Growth is derived partly from its electronic trading capabilities, the filing noted.
Mathisson's and Santayana's efforts may have helped to save CSFB's equities group from financial oblivion, but their teaming certainly didn't result from any five-or ten-year plan. AES came together in an almost ad hoc fashion.
In 2001, when penny ticks changed stock trading forever, Mathisson and Santayana were working at different ends of CSFB's electronic trading world. Mathisson, responsible for development of AES's algorithms, was a trader on CSFB's proprietary desk. He had joined CSFB in 2000 as a statistical arbitrage trader after spending eight years with famed quant house D.E. Shaw.
Santayana, in charge of sales and marketing for AES, was working in sales and marketing for CSFB's FIX connectivity services in 2001. He joined CSFB in 2000 as an employee of Donaldson, Lufkin & Jenrette when that firm was bought by CSFB. Santayana had previously worked at State Street Global Advisors marketing its Lattice electronic trading product. He was also in sales at IBM. What brought the two men together was a cry for help from the buyside. Decimalization had caused a sharp decrease in visible liquidity. That made it difficult for institutional traders to trade large blocks.
"People couldn't get their trades done," Mathisson said. "Guys who had been trading for 20 years were unsure of the best way to move orders around. The sizes in which they had always been trading were all of a sudden too large. So, there was a huge need coming from sales traders and buyside clients."
Mathisson and his colleagues on the prop desk encountered the same problems. Using their knowledge of quantitative methodologies and automated trading, the desk started to build systems that would break up large blocks and feed the orders into the market in piecemeal fashion. At CSFB, the algorithm was born.
With the algorithms developed, the next problem was distribution. Should CSFB build or buy a front-end trading system, install it on hundreds of clients' desktops, and then deliver the algorithms through the proprietary technology? Or should it find an easier, faster way?
That was where Santayana and his FIX group came in. Rather than go to all the trouble of commandeering and promoting a front-end (as did Morgan Stanley with Passport; Goldman with REDI; or ITG with QuantEx), CSFB decided to cut deals with vendors already on buyside desktops.
"That made it incredibly easy for the clients to use the product," explained Santayana. "They could use whatever product was already on their desk. They didn't have to learn a new system. They didn't have to change their trading behavior."
The first system and the key to AES' early success was Bloomberg. CSFB signed a contract with Bloomberg that gave them exclusive representation for two years. Now, Bloomberg has about a dozen algorithm providers on its ubiquitous market data and analytics terminal. But, for the first two years of the algo revolution, there was only CSFB.
CSFB is now embedded on nearly 40 vendors' platforms. These include all of the major order management and front-end trading systems. All have some sort of revenue-sharing arrangement with CSFB. "We cut deals quickly and aggressively at first," noted Mathisson. "We were very quickly on all the major platforms." The firm recently hired an exec just to manage those relationships.
Access to AES' algorithms by CSFB's customers is now done exclusively through those third-party platforms, according to Mathisson and Santayana. None of the orders come in through sales traders.
AES has a sales force of about a dozen people who call on the same clients as CSFB's sales traders. The set-up contrasts sharply with those of some of CSFB's competitors. Lehman and JP Morgan, for instance, market their low-cost algo trades through their sales traders. The idea is to avoid internal infighting and present a single face to the client.
At CSFB, the AES executives play down any friction between their team and CSFB's sales traders. "Senior management got behind the product in 2001," notes Mathisson. "They realized where the market was headed." Santayana added: "There are no factions of competing interest."
Senior management has supported electronic trading at CSFB, say sources close to the firm, but the deployment has not been frictionless. Some sales traders objected to the inherent cannibalization of the process. In other words, trades that they might have done at 5 cents were now being done by a machine at 1 cent.
AES is a "no-touch" service, meaning the algorithms are accessed by the buyside without human aid. That kind of business is generally done on the Street for about 1 cent per share. Trades done on a full-service basis generally earn firms about 5 cents per share.
The AES execs maintain that most of AES' order flow has been incremental. That means it has been won from competitors or through general growth in market volume-not from other CSFB desks. CSFB's AutEx BlockData numbers, however, show a slight decline in CSFB's overall volume in the past three years.
If it is true that CSFB has been grabbing share from its competitors, then it is equally true that those competitors want their shares back. Despite the lead CSFB (and Goldman and Morgan Stanley) has over rivals, they are catching up.
Banc of America Securities, Lehman, Citigroup, UBS, Merrill Lynch, and JP Morgan make up the second tier of bulge-bracket shops with electronic ambitions. All are spending heavily to ensure they have offerings comparable to the Big Three.
Mid-sized firms such as Piper Jaffray, ITG, Instinet, Sanford C. Bernstein and a host of tiny, specialized electronic shops are also plying the algorithmic waters with proprietary technology.
More than one study into algorithmic trading, though, has concluded most brokers' algo offerings are too similar. "Algorithmic trading is pervasive," states a recent report from Financial Insights, "but undifferentiated among bulge-bracket firms."
The consultancy based its conclusion on two observations. First, there are now many brokers offering many algorithms. Second, the presence of algorithms at a brokerage doesn't influence the buyside trader's choice of brokers to a very high degree. "Differentiation is difficult," the report concludes.
CSFB however doesn't seem to be listening. "If you don't have algorithms," says Mathisson. "If you don't have good algorithms, you're not participating in a big chunk of the flow. You're just not going to get it."
CSFB'S Crystal Ball
On the pre-trade cost analysis front, CSFB will soon launch ESP, which stands for Empirical Slippage Predictor. Developed by in-house quant Merrell Hora, the technology forecasts the cost of a potential trade. As with all pre-trade programs, ESP bases its decisions on current market conditions and historic trade information. Where it differs is in the source of its historic data.
Rather than use historic NYSE and Nasdaq price information, ESP bases its decisions on a scan of all of the trades CSFB's AES desk has done in the past four years. CSFB claims this methodology will provide more accurate forecasts. "We have tens of millions of trades in our database," Mathisson notes.
Basket Cases
Also on the algo front, CSFB has launched PHD, for portfolio hedging device. The program is geared toward portfolio traders who want to place restraints, or limitations, on their executions at both the portfolio and individual stock levels.
There are systems on the market that let traders "optimize," or place constraints, at the portfolio level. There are also systems that do the same at the individual stock level. But there are none, according to CSFB, that do both.
As an example, assume a basket of 500 stocks. With PHD, the trader can set constraints on individual names that restrict their trading to perhaps no more than 10 percent of the volume. Or they can use PHD to track the VWAP. At the same time, the trader can enter constraints at the portfolio level that might minimize tracking error vis-a-vis the S&P 500 index.
Guerilla Tactic
Algorithms, for the most part, have been used to trade large-cap names. They have not been as effective with the stocks of smaller companies. CSFB is introducing an algorithm called Guerilla. AES claims it maintains a new invisible presence in the marketplace.
Guerilla has access not only to the traditional market centers such as the New York Stock Exchange and Nasdaq, but also CSFB's own Crossfinder internal crossing system and a collection of alternative trading systems.
The ATSs are used primarily by institutional desks to trade small-and mid-cap names. CSFB won't disclose the names of the third-party systems to which it connects. The majors, though, are POSIT, NYFIX Millennium, Lava Dark Book, Pipeline and Instinet's CBX.
Guerilla essentially does two things. It lurks out of sight within the various trading venues, all the while waiting for liquidity to present itself. When an agreeable bid or offer does present itself, Guerilla strikes, hitting the bid or lifting the offer.
"The math behind it is very difficult," notes Mathisson. "It is looking for trails that other traders are leaving behind." In other words, it is looking for signals that there are sells in the market.
Second, Guerilla covers its tracks. Once it trades, it disappears. It could be gone for three minutes, an hour or the rest of the day. "Nobody knows you are out there," Santayana says.
Three CSFB quants, two in New York and one in London, worked for two years on the math behind Guerilla, according to AES execs.
