Carl Carrie, JP Morgan:
We believe reverse engineering of algorithms is happening in the marketplace. Our process for minimizing the risks of this include algorithmic randomization of venue, size, wait time, limit prices and micro-order types. In general, algorithms should avoid pegging between the NBBO predictably and instead use non-posting order types and venues appropriately. We avoid trying to be predictably reactive (e.g. too large quotes, narrow spreads) and use other stealth techniques to confuse arbitrageurs and minimize the profit potential of algorithmic gamers.
Frank Brown, EdgeTrade:
Without question, reverse engineering is taking place both upstairs and on the floor. More egregious, is the reverse engineering carried out by the brokers with proprietary trading desks to whom buy-side firms are sending their algorithmic orders. EdgeTrade's independent agency model makes reverse engineering by us, the broker, a non-issue. We also go to great lengths to design algorithms that blend their trading into the overall market activity, reduce the footprint our clients' trades leave in the marketplace, confound attempts at reverse engineering and minimize market impact and slippage.
Q: How important is pre-trade analysis when using algorithms? Do you see that importance of pre-trade analysis growing? Describe your pre-trade analysis tool and how it gives investors an edge?
Richard Johnson, Miletus Trading
As a quantitative brokerage firm, we believe pre-trade is a very important tool for traders, particularly when it comes to determining appropriate trading strategy. However, it is important not to look at pre-trade in isolation from other components of the trading process: pre-trade only offers a static snapshot based on historical data, and the recommended optimal trading strategy may no longer be the most appropriate given changing market conditions.
At Miletus, our market impact model is integrated into our algorithms and our Strategy Interface. Thus, when users determine their desired aggressiveness and trading horizon using our Trade Impact Estimator, they can input these parameters directly into our algorithms. This means the algorithm can intelligently adapt to changing market conditions as opposed to doggedly following an outdated pre-trade strategy.
We are definitely seeing increased demand for pre-trade analytics, especially from portfolio managers who are using our tool in the portfolio construction process.
David Liles, Bernstein:
Pre-trade analysis tools help the algorithmic trader by setting reasonable expectations regarding trading costs, identifying outliers, and assisting in the choice of an appropriate strategy. Where pre-trade analysis truly shines is when it is used in conjunction with portfolio-level algorithms. Bernstein's pre-trade analysis offers clients detailed estimates of trading costs and confidence intervals for multiple implementation scenarios including our PortFall algorithm, trading each stock independently using Price Capture (our single-stock IS strategy), and also using a VWAP-based strategy. We assist clients in handling outliers by highlighting stocks that may be ill-suited for automated strategies due to their individual trading characteristics. In a portfolio setting, we also alert clients to the potential effect that removing a single name for manual handling will have on the overall risk of the portfolio.
Derek Morris, BNY Brokerage:
There is continued debate about the effectiveness of pre-trade analysis, as some believe that it offers few advantages in being able to increase trading performance. Yet, pre-trade analysis is a helpful tool in understanding all the costs associated with trading. Pre-trade has evolved rapidly beyond mere volume impact predictions. Real-time products have begun to come online. If liquidity remains fragmented, pre-trade analysis will have a place for a long time to come.
We use FlexTrade's real-time TQM product because it is the perfect complement to FlexTrader, which we use on our Program Trading Desk. With TQM, we can pinpoint the source of our cost and determine true market impact in real-time and historically.
David Mortimer, Piper Jaffray:
Pre-trade analysis is a very important tool for selecting an appropriate benchmark-and the execution strategy, time horizon and trading parameters to meet that benchmark. The determination of alpha is the single most important part of this measurement-unfortunately, with the products currently being offered to the buyside-predicting alpha is still more art than science. Through an association with Quantitative Services Group (QSG), APT has recently introduced a pre-trade product that will move the ball forward, in terms of the accuracy of pre-trade market impact and opportunity cost measurement. For example, our forecast tool uses extensive empirical data in addition to the commonly used factors-volatility and spread size. For portfolio trades a dynamic covariance matrix is employed to take into account symbol and industry correlations. Additionally, since our models are accessible over a secure web interface, and cannot be seen by APT or any broker-dealer (QSG hosts the site), our clients are not at risk of information leakage while use our pre-trade analysis tools.
Andrew F. Silverman, Goldman Sachs:
The proliferation of algorithms can be perplexing: clients are asking for advice on how to choose the right algorithm? And within each algorithm: How to optimize the various parameters? Pre-trade analysis is extremely important in answering both these questions. At Goldman Sachs we developed the Cube pre-trade framework to classify orders by difficulty and map them into the appropriate execution strategy and algorithm. Moreover, as part of our suite of pre-trade analytics, our GUIDE tool uses our cutting-edge trading cost model to give clients pre-trade estimates for different execution aggressiveness and help "guide" them to the strategy that best meets their needs.
John Wightkin, QSG:
Pre-trade analysis is very important when using algorithms (or any kind of trading). In addition to predicting the expected cost, a good Pre-Trade system should help traders identify the most appropriate/optimal trading strategy based upon the goal they are trying to achieve.
Pre-trade analysis will grow regardless of the adoption rate of algorithms. As firms begin to look for competitive advantages in other parts of their investment process, the ability to improve their trading decisions will become more critical.
Our pre-trade model is unique in that it incorporates both our tick based metrics and our stock selection research. These two inputs allow us to provide our clients a more accurate estimate of their total trading costs (market impact plus a directional forecast). This estimate also allows us create a trading algorithm that best suits (minimize the trading cost, and hence maximize the realized return) each firm's unique investment process.
Jarrod Yuster, Merrill Lynch:
How important is pre-trade analysis when using algorithms? DO you see that importance of pre-trade analysis growing? Describe your pre-trade analysis tool and how it gives investors an edge.
The ML X-ACT team recognizes the growing importance of pre-trade analysis. Our "edge" lies in the extensive databases fueling our pre-trade offerings. We offer pre-trade guidance on both the single stock and portfolio level via products including algorithmic criteria checks, algorithm chooser and GEA, an equity analytics application which provides estimated impact cost, liquidation scenarios and risk analytics for global portfolios.
Frank Brown, EdgeTrade:
Pre-trade analysis remains useful in helping a trader better understand the characteristics of a trade-whether or not it is to be executed algorithmically-although no more so than it was when it first became a packaged product a decade or more ago. We do, however, envision pre-trade analysis increasing in usage, particularly as focus on best execution deepens and algorithmic trading gains broader acceptance. EdgeTrade views pre-trade analysis as much more than a pre-packaged tool. Through our hands-on, consultative approach with clients, we work with traders to construct customized tools that help them trade a choice of styles-adapted to their specific trading goals-more effectively and efficiently.
Carl Carrie, JP Morgan:
In the context of algorithmic trading, our clients are looking for differentiation in the marketplace-tools that can help them achieve their macro-goals and help them better predict their costs upfront.
GES' pre-trade analysis uniquely provides measures for trading difficulty, volume volatility, volume profile variance and cost estimation. These measures are directly attributed to specific algorithmic strategies. Pre-trade estimates such as these reduce the probability of incorrectly using algorithms and align investor goals with actual implemented results. Our pre-trade analytics is now directly embedded in many of our single stock algorithms (e.g. Implementation Shortfall) and portfolio algorithms (Trading Algorithmic Optimizer-TAO). In 2006, we expect the use of pre trade to grow as the basis for optimal algorithmic trading.
John Coulter, Vhayu:
The importance of pre-trade analysis sometimes gets overlooked in the algorithmic trading chain. There are two different components to pre-trade analysis that we see from our clients today; finding opportunities and conducting impact cost analysis before implementing a strategy. Many of our customers develop and test complex trading strategies for internal use which then get pushed out to institutional fund managers. The broker's applications search for real-time opportunities and automatically apply an appropriate algorithm based on pre-trade decision support tools. Some customers build tools on our platform to measure the results versus various calculated benchmarks to enable concrete validation. Our solution combines real-time equities data, historical TAQ data and internal reference data on the same platform to support any front-end application. It's all done on a single server that has built in fault tolerance. Normally these solutions are developed, tested and in production in less than 3 months.
Brian Fagen, Morgan Stanley:
We believe that pre trade analysis is the key for all equity trading, not just electronic trading. The true value of a pre-trade tool is to give the trader information to aid in the decision making process of how to trade an order. At the first level, this decision is between using capital, working on the floor or with a market maker, looking for a natural cross, or trading electronically. The next level would be at the electronic trading level, looking at which algorithm might be best, or what other potential strategy would deliver the most optimal execution.
Our pre-trade analysis tool- BXS Navigator-gives the trader full transparency into estimated impact cost and risk of a particular order depending upon the trading strategy chosen. Navigator allows the trader to perform analysis of different trading strategies to determine their expected cost prior to trading. This helps the trader determine the best possible strategy based upon their investment objective.
Tony Huck, ITG:
ITG has a unique position as the provider of ITG LogicTM, our advanced pre-trade analysis tool. ITG Algorithms leverage this leadership by providing integrated algorithmic trading solutions to carry out the results of pre-trade analysis. Several large clients have taken this a step further, closing the performance feedback loop by employing ITG's post-trade tool, TCA, to continually enhance their trading strategies. The ability to examine vast amounts of trading data to provide optimal execution strategies is growing in importance as trading becomes more automated and traders become more sophisticated.
Q: What specifically about your technology or team leads you to believe that your firm will be a long term player in the algorithmic space?
Tony Huck, ITG:
Any firm hoping to be a long-term player in the algorithmic space must continually innovate and enhance performance. Ultimately, those who survive will be the ones that provide the most value to their customers by improving their clients' trading outcomes and thereby improving their clients' profitability. Combined with the trust engendered by its agency status, ITG is ideally placed with its unified expertise across the spectrum of trading, from trade order management, pre-trade analysis, execution front-ends and algorithms, and post-trade measurement. This unique range of capabilities allows ITG to provide more integrated and innovative solutions to its clients.
Brian Fagen, Morgan Stanley:
Morgan Stanley has developed a world-class team throughout our electronic trading business, including financial engineers, econometricians, technology & software developers and client focused sales and service. This team is focused on developing trading solutions that add value to our client's investment performance. This is evident in our pre and post trade tools, the stability of our robust infrastructure, our algorithmic trading strategies, the Passport front-end system, and the quality of our people who provide service and consultancy to our end clients.
We believe that the true differentiators in this business will be a firm's ability to have a meaningful value added to their clients business across the entire platform, globally, and across asset classes.
John Coulter, Vhayu:
Market data volumes have been increasing at an astounding rate. In 2004 alone, data volumes grew by over 100%. That trend is expected to continue. These volume increases pose significant technical and financial challenges for market data vendors and their clients as they race to put the necessary systems in place to efficiently cope with the data volumes. Not only has the sheer volume of data skyrocketed, but its efficient use and distribution has never been more critical to the profitability of brokers. The only guaranteed way of supporting the impending electronic trading paradigm is to deploy data servers capable of processing ticks hundreds of times faster than just a few years ago with virtually no latency. This effort is now a necessity rather than a luxury and we can be certain that brokers and institutions will maintain a high priority for tick processing and persistence technology.
Carl Carrie, JP Morgan:
Our team is uniquely positioned to innovate and become a leading player in the algorithmic space with the investments in technology and financial engineering that we have made. We believe we offer the highest quality and deepest breadth of product in the industry. With our recent acquisition of Neovest, we offer a truly broker neutral, multi-asset class global execution management system for algorithms and DMA. Our portfolio algorithmic product, TAO, is an industry 1st for integrated analytics, optimization and portfolio trading. Our extensive tick database and analytical engines are global and based on some of the most sophisticated technologies and financial engineering available. Finally, our algorithmic and quantitative product development effort is supported by a single and completely dedicated team of world-class transaction cost analysis experts and gifted quantitative researchers.
Frank Brown, EdgeTrade:
EdgeTrade's philosophy and practice is to have algorithms designed by those with real-world trading experience. Our quantitative team understands what it takes to create the right tool for a particular trade. Our technology team delivers fine-grained control of these algorithms through Execution Management System-EdgeTrade's independent, agency front-end platform for trading single stock and basket equity orders-or integrated with industry leading OMS and third-party trading systems. EdgeTrade's conflict-free agency environment also means we can design tools for traders that other brokers with competing fiefdoms are unwilling to offer. Our strength is in the independent agency business model that enables traders to efficiently and anonymously execute trades using algorithmic and DMA tools developed for practical, real-world purposes.
Jarrod Yuster, Merrill Lynch:
The full backing and commitment of Merrill Lynch, one of the world's most resource-rich firms, to the ML X-ACT algorithmic trading platform provides for a seamless global offering, grounded in a rich data infrastructure. Because this platform was built with flexibility in mind, ML X-ACT can facilitate a range of customizations down to the individual trader level. Additionally, Merrill's expertise extends beyond electronic trading, allowing for execution guidance and consulting across block and program trading, capital commitment, and electronic trading.
John Wightkin, QSG:
Being one of the few independent providers of trading cost research, our objective analysis will always be needed to critically evaluate and compare algorithms. Also, our forensic investigation, coupled with our unique measures, provides the marketplace insights into the algorithms that are difficult to find from any other source. As the use of algorithms grow, the demand for better transparency, monitoring, measuring and analysis will follow. This demand should ensure our longevity in this space.
Andrew F. Silverman, Goldman Sachs:
We are committed to providing advanced global algorithmic trading tools for a multitude of financial instruments for both internal and external users at Goldman. Every trading desk within Goldman Sachs either has or will have the same advanced platform that we are offering our clients. As a result there is a tremendous amount of vested interest in making this platform be the best available. Our product development team has an open and constant dialog with a diverse range of GS's internal trading desks searching for new algorithmic ideas, feedback on performance and functionality, and robust testing. Our primary focus is to work very closely with our clients, to better understand their needs and to build the algorithms that our clients think will improve their ability to execute effectively. But having access to internal traders during our product development process, as well as having internal and external users utilizing the same platform, allows Goldman to concentrate its resources and efforts on developing a single set of highly competitive algorithmic execution tools.
David Mortimer, Piper Jaffray:
The answer to this is simple:
Advanced trading technology built by folks whose only interest is in achieving best execution for our clients;
We have been a market leader in the introduction of unique and differentiated products such as FUSION, our price-predictive ESP algorithm, and our pre-trade analysis models;
Our quant team has been exclusively developed in-house-with no outside Wall Street experience or "industry" information. This ensures that we look at every problem from our own unique vantage point; NO PROP TRADING.
Derek Morris, BNY Brokerage:
We are well positioned to anticipate and respond to our clients' needs in this fast changing landscape. We have invested heavily in technology to deliver the tools our clients' need, whether it is buying, building, or partnering with best-of-breed providers. Our clients have come to rely on us to deliver full-service capabilities without the conflicts. Our long-term commitment to a pure agency model positions us as a trusted partner for our clients.
We have an accomplished and dedicated team of experts. The market intelligence they gather, by executing on behalf of our clients, is a significant component of our algorithmic development. Effective algorithms result from a willingness to adapt human input to technology advances. We are always prepared to reinvent our development process to drive incremental performance from the marketplace. The algorithm is not the end game; it is one more necessary tool for executing in today's market.
David Liles, Bernstein:
The single most important reason that Bernstein will remain a long term player in the algorithmic space is the quality of the product. We have built our product offering around the latest academic research, often from beyond the field of finance, and have used best of breed technology in the infrastructure. Another critical factor is the Bernstein business model. As an agency-only shop with no proprietary trading whatsoever, we are able to offer algorithmic clients true anonymity. Additionally, there is a high level of cooperation between the cash and algorithmic desks at Bernstein such that clients get the same positive message from both teams. Last, but certainly not least, the top-tier research product offered by Bernstein puts a huge wind at our backs. All of these factors combine to make the Bernstein algorithmic trading offering a compelling choice for the buy side trader who wants a top notch agency execution at a firm where their commission dollars will truly be put to good use.
Richard Johnson, Miletus Trading
Our goal is to establish Miletus as the premier algorithmic brokerage firm. Miletus Trading was founded on the belief that the buy-side needed more efficient user friendly trading tools to establish real-time quantitative execution strategies. We have assembled an elite Research and Development staff with backgrounds in mathematics, computer science, electronic trading systems and quantitative analysis. Our cutting edge technology infrastructure, built from scratch less than two years ago, is optimally designed for algorithmic trading. Our mathematical foundation yields models that need to be implemented in a logical and intuitive manner. In order to achieve superior performance, we have devised a unique formula that combines our technological infrastructure, mathematical modeling and trading expertise. As a result, our clients have greatly benefited from our highly specialized method of algorithmic trading.
