Algorithmic Trading: Goldman's Piccolo algo

Discussion in 'Metal Futures' started by stephencrowley, Jan 10, 2006.

  1. Anyone know much about this yet? I have access to it thru my broker and was wondering if it is worth using.

    Goldman Sachs Jolts Algo Trading Suite

    With an updated algorithmic trading suite, Goldman Sachs targets faster and more profitable trades.

    Goldman Sachs jolted its algorithmic trading suite last month with updated algorithms and a new capability that officials say will help traders find the best algorithms for their orders. The new feature for Goldman's RediPlus direct market access (DMA) platform is dubbed Navigator. In addition, the investment giant is working on a number of enhancements to its Goldman Sachs Algorithmic Trading (GSAT) set of offerings.

    As reported in Dealing with Technology, Goldman officials foresee algorithmic trading as increasingly applicable to other asset classes, and are planning to develop algorithms that look at more than one security or a single asset class. Secondly, greater levels of customization and control are being added to GSAT's offerings as clients seek to design their own algorithms and determine parameters for how each algorithm works their orders.

    Navigator aims to help traders choose the most suitable algorithm for any given order. The RediPlus order execution and quotes software, generated by the Spear, Leeds & Kellogg (SL&K) unit of Goldman Sachs, provides access to liquidity for US equities, options and electronic futures as well as European equities and electronic futures markets.

    Greg Tusar, head of electronic execution product development, says that the decision to build Navigator was driven by the growing number of offerings hitting the market. "We've been getting a lot of questions around, 'How do I know which algorithms to use under different circumstances?' In creating Navigator, we've created an algorithm for the algorithms. You give the order to us and Navigator will decide which algorithm will be most effective to execute that order," says Tusar.

    The market is taking off for algorithmic offerings. Analysts at market researcher TowerGroup estimate that 70 percent of orders from the buy side to the sell side are executed via electronic direct market access, algorithmic and program trading channels while 30 percent of orders are directed to block trading desks. "The buy-side firms are gravitating more to the brokers who have a complete suite of execution services," says Dushyant Shahrawat, a TowerGroup analyst.

    Navigator will have the ability to decide between two of Goldmans' algorithms—Piccolo and Forecast. However, the long-term goal is to increase the number of inputs to allow greater sophistication in Navigator's decision-making process. "Going forward, we'd like to make it robust enough to look at all of our different kinds of algorithms," says Jana Hale, global head of GSAT.

    Piccolo is a spread capture algorithm designed for small order sizes. In order to capture some of the spread, the algorithm splits any given order, sending part of it as a market order and the remainder as a limit order. While the market order is executed immediately, the limit order will stay on the order book to capture some of the spread. But if the limit order doesn't execute within a given time frame, it will convert to a market order.

    With any given algorithm, different clients may look to use different parameters on how that algorithm behaves. Using the example of Piccolo, Tusar explains that, "some customers might wait for five minutes for the limit order to be executed, others might wait for 15, while others may only want to wait for a minute." Piccolo can be customized on a per-order, per-day or per-client basis.

    In addition to changing parameters within an algorithm, Michael Towarek, head of GSAT Europe, says that some clients may also request customized or internally developed algorithms. "We're looking more and more at totally custom-made tools for some of our top clients," he says.

    However, the firm must always look to balance revenue opportunities with client service because responding to every request for customization may not be economically viable, especially given the thin margins in algorithmic trading.

    "We're trying to create building blocks of algorithms that can somehow be adapted, changed and attached to each other. Once you create that basic menu, customization doesn't have to be extreme. It can be adding different elements or rules that are built into one or two algorithms and modifying them slightly," says Hale.

    The small army behind Goldman's algo efforts has one major goal—to provide a desktop application "where clients can trade multi-asset classes using trading algorithms across any of those classes," says Andrew Silverman, head of GSAT US.

    Multiple asset-class platforms may inevitably require multiple asset-class trading algorithms. However, the exact nature of these algorithms has yet to be determined. Hale says she foresees multiple asset algorithms such as "doing ratio algorithms between companies, or down the line, a convertible versus an equity. I think that's how things will progress."

    As a result, Hale says that algorithms are going to become much more complicated than the average, and commonly preferred, VWAP benchmark.

    Given the success of algorithmic trading over the last couple of years, could the industry be in line for a future dominated by robot traders? Hale says she thinks not, principally because algorithmic trading cannot access hidden liquidity. "Through algorithms you can only ever address exposed liquidity in marketplaces. There will always be a demand for sophisticated advice in how to bring buyers and sellers together," she says. "Big blocks of stock will not always trade in certain manners, and certain floats and small cap stocks are not going to be easily addressed by algorithms."
  2. If I remember correctly, Piccolo is just a block book (aggregating all the GS internal and client flows) algorithm, at least that's what I heard at the automated trading workshop that GS sponsored around beginning of '05.

    Let me see if I can dig up the presentation somewhere (it is probably in a heap of buried paper).