Jeffrey Hogan Comments: Algos Take Hold in Fixed-Income Markets- Securities Industry News

Algos Take Hold in Fixed-Income Markets

Want to make $10 million a year? Simple: Be able to create algorithmic, high-frequency trading applications for bonds and fixed income derivatives that make your firm lots of money.

If your high-speed apps can find arbitrage (read: lucre-making) opportunities, such as being able to pinpoint real-time pricing discrepancies between cash and bond futures, say, or quickly hit and lift optimal Treasury prices streaming from competing inter-dealer brokers' electronic bond trading platforms, you've got a shot at an eight-figure income.

Anyone unconvinced that algorithmic trading has come whole hog into the fixed income world should look at the want ads. Take this posting-and there are many like it-from Jan. 11: "Our client seeks... automated analytical black box trading strategists with proven track records at top-tier investment bank, hedge fund or prop shop ... high-frequency trading strategies for fixed income cash bonds, interest rates and credit derivatives. Compensation: 1M-10MM."

Yet, until recently, algorithmic trading wasn't much applied to fixed income, and then, only sparingly in basic VWAP (volume-weighted average price) analyses. VWAP is a passive strategy aimed at trading at prices that basically match the amount of trading occurring in the security at the time, so the market impact of one's own execution is as slight as possible.

Ubiquitous in equities for years, algos for bonds are now much more complex, varied and numerous, as certain asset classes have become more electronic and therefore applicable: What's made this possible are streaming quotes from the e-trading venues of inter-dealer brokers, which have provided the foundation for automated, model-triggered strategies. Without live, streaming quotes, similar to those long available in stocks on equities exchanges, algorithmic trading is difficult to nigh impossible.

"The high transaction volumes taking place between a relatively small and well-defined set of counterparties make IDB [inter-dealer broker] markets the logical front line for fixed income algo trading," says Billy Hult, president of Tradeweb, which since 1998 has run a multi-dealer-to-buy-side electronic bond trading platform but last year launched an inter-dealer platform called Dealerweb.

"In the 18 months since we've entered this business with Dealerweb ... we've seen an increasing trend towards more automated trading," Hult said.

SPECIFIC BOND ALGOS

Algo-based bond strategies include but are not limited to trading on price discrepancies between Treasury cash bonds offered on the major IDB platforms and Treasury futures on futures exchanges. Similar "arbitrage" strategies are increasingly being applied to interest rate and credit default swaps, as more of these derivatives are trading electronically ahead of regulatory overhauls calling for same. Many current algo job postings are asking for expertise in these over-the-counter derivatives.

A lot of the activity so far, though, occurs over the futures exchanges that offer bond derivatives, like the CME Group, Eurex and NYSE Euronext.

"What we see most in the algo space in Europe involve interest rate futures, Eurodollar futures and bond futures," said Steve Wilcockson, industry manager of computational finance at The MathWorks. The Natick, Mass.-based firm's Matlab programming language is widely used by financial firms to construct proprietary algorithms and quant applications.

"Quite frequently hedge funds will also construct their own yield curves to compare and contrast with the yield curves the market data vendors provide." The idea here is to act quickly on what may be incorrect-albeit consensus-pricing from vendors known for selling "de facto" yield curve data, such as that from Bloomberg or Markit.

Other advanced algorithmic models include multi-legged trades, some based on price comparisons between bonds in several different currencies with the values of the currencies themselves and what are essentially over-the-counter futures, called forward contracts, on those same currencies.

The duration of some of these trading strategies can occur over the course of a day, though, so they are more "intraday" than sub-millisecond: "High frequency" as applied to such bond algos would be a misnomer. But using a database like Kdb+, from Palo Alto-based Kx Systems, co-located at the data center of an electronic platform that offers streaming bond data, combined with a historical in-house archive and well-written models set with parameters to trigger immediate trading can lead to automated, split-second decision-making at the fixed income desk. Kdb+ is a time-series database used by Cantor Fitzgerald, Goldman Sachs, J.P. Morgan, Morgan Stanley and Fidelity Investments to make real-time trading decisions across asset classes; it's written in "Q," Kx's own programming language.

"Complex event processing is getting more complex, because the banks are looking at more data in more complicated ways," says Simon Garland, the chief strategist at Kx. "We power that, if you really want to get top performance across vast amounts of data."

The open source "R" programming language, which can be used with finance building applications from The MathWorks, is lately giving Kx a run for its money; rival database provider Los Gatos, Calif.-based Vhayu Technologies, offers R integration, for example.

While fixed income algo trading remains primarily "focused on the inter-dealer space," Tradeweb's Hult says, model-triggered bond trading is beginning to expand outside of the realm in which only the sell-side dealers play.

"There is a lot of top talent available because of the financial crisis," Hult added. "Good quants have been in demand for years; on the buy-side, hedge funds have always been at the forefront of trading innovation. But we are increasingly seeing traditional asset managers focusing on their use of technology to ensure better execution."

Note, however, that buy-side bond execution remains predominantly request-for-quote (RFQ), in which asset managers must request a price quote from one or several sell-side shops. This is a less efficient and less competitive trading model over which high-frequency, algo-based trading cannot readily be applied. This compares to streaming quotes, in which pricing is continually updated from live bids and offers visible to all dealer-participants. The latter inter-dealer, or "wholesale," market model, where dealers trade among themselves, is how stock exchanges work, essentially.

However, in the last several years, shops big and active enough to break the mold of what's traditionally been considered "buy-side" and "sell-side," like hedge fund giant Citadel Investment Group, for instance, have been allowed to participate on inter-dealer bond platforms. Their high-frequency trading has enabled them to smash the taboo the dealers-and hence the IDBs-once had against allowing firms generally considered "buy-side" into the wholesale markets. IDBs offer sponsored access as well, in which a buy-side shop chooses a primary dealer, or several, through which they can trade on the IDB platforms.

Now Citadel and GETCO, a Chicago-based high-frequency model-driven market maker, are among the top players in algo bond trading over the platforms. GETCO, for instance, which did not respond to questions regarding its algo bond trading, claims it trades volumes that rank it among the top five participants over Icap's BrokerTec and BGC Partners' eSpeed Treasury platforms. Some shops' fixed income trading now is almost completely automated and model-triggered, and only manually overridden during times of volatility, which is what the bond fund of Paris-based hedge fund Rivoli Fund Management does.

WHERE THE ACTION IS

By far, the most algorithmic trading in fixed income currently occurs in the most liquid bond sectors-namely U.S. Treasuries, where both Brokertec and eSpeed dominate electronic market share. Dealers use their own algos over these platforms.

Algorithm-based trading has grown to now account for 45 percent of overall trading in U.S. Treasuries over Brokertec, said Icap COO Mark Yallop in November, during a report on the firm's half-year earnings.

Yallop said that "model" or algorithmic trading of foreign exchange (FX) now accounts for more than 50 percent of daily volume on the firm's EBS Prime platform, and 35 percent of volume in overall FX trading. EBS Prime and Multi-Prime Access offers buy- or sell-side users live executable prices from prime brokers. EBS Spot Ai offers banks and hedge funds automated, algorithmic trading in FX.

He added that 275 customers of its EBS Prime FX system and 54 customers of its EBS Live platform use Icap's i-Cross, a co-location solution which places firms' servers alongside Icap's servers in its data center to speed up the customers' trading and reduce latency in their models.

Icap did not respond directly to questions about the trends and type of algorithmic trading done over Brokertec and EBS, except to email boilerplate language from its Web site describing i-Cross. Icap's co-location solution offers close server proximity for U.S. Treasuries in North America and for FX trading in North America, London and Tokyo.

BGC would not speak directly about algorithmic bond trading occurring over eSpeed either. Both Icap and BGC took pains in emailed statements to emphasize that they provide fair access to customers. Regulators have recently frowned on automated strategies that include flash trading, primarily in equities, though.

Regardless, Jeffrey Hogan, BGC's managing director of business development, offered a word of caution regarding algorithmic bond trading, by mapping out what he called a "long-term and non-myopic view" of how electronic bond venues and their users might manage technology to benefit the overall marketplace.

"We believe the full spectrum of users need to be able to jump into the pool together, from the most lightning-fast price-taker, to the point-and -click arcade member, to the screen-assisted voice broker, and the purely voice-supported client. The high-velocity companies running everyone else over is not positive for their own long-term viability. Therefore, we are devoted to create sustainable hybrid voice and electronic marketplaces, which do not especially advantage the high-velocity participant."

Back to press releases landing page