One of the hottest topics in the futures markets these days is algorithmic trading. The financial press is awash with story after story about the rise of algorithmic trading, and a trader would be hard-pressed to attend a futures industry conference that did not have multiple panels devoted to the topic. However, the content of most of these articles and the substance of the conference chatter typically center on how algorithmic trading has been readily adopted by a new breed of traders — the so-called “quants” — who are utilizing ultra-high-velocity, short-term algorithmic strategies that are designed to identify, react to and profit from the slightest market inefficiency, all in a matter of milliseconds.
While the growing impact of the quants and “the need for speed” are important stories with significant implications for the way futures are traded today (and tomorrow), what’s been missing in the discussion is why most commodity trading advisors (CTA), who as a group are among the most active of futures traders, have been slow to adopt algorithmic trading tools, and how the adoption of such tools can help them improve trading performance; strengthen risk management; and automate many middle- and back-office functions, such as allocations and give-ups.
Algo Trading: What Does It Mean?
Before exploring how CTAs can benefit from algorithmic trading, it’s necessary to define the term. “Algo” trading is used very loosely, often as a catchphrase for any or all of the following: black-box trading, computer trading, automated trading, systematic trading, program trading, basket trading and enhanced execution. This is not an exhaustive list. At its core, algo trading is simply rules-based trading. But a key distinction within algorithmic trading must be made.
An algorithm can be the trading strategy itself — the rules that determine which market to trade and when to trade it. For example, a strategy that employs a set of technical indicators, such as Moving Average Convergence/Divergence (MACD) or On Balance Volume (OBV), is algorithmic. Such “strategy” algorithms are at the core of most CTA trading systems.
But an algorithm also can be employed to execute an order in a manner that will reduce market impact and slippage, thus improving trading performance. In such an instance, and in the discussion that follows, the term “algo trading” is interchangeable with algorithmic execution. Examples of “execution” algorithms are Time Weighted Average Price (TWAP), which breaks up a large order and executes it evenly over a set period, and Volume Participation, in which each piece of a large order is executed in line with overall traded volume.
CTAs and Execution — Past and Present
When defined as a rules-based system for execution, algo trading would seem a natural fit for most CTAs, who typically employ a very mechanical and systematic approach to all their work processes, from identifying alpha opportunities to managing risk, compliance and account allocations as well as back-office operations. And yet, most CTAs, especially those who employ trend-following or other systematic trading methodologies with intermediate- to longer-term time horizons, have been slow to embrace algo trading. Why? The reasons have largely to do with the nature of the traditional CTA business model and that many key futures markets have not provided, until very recently, a liquid electronic trading arena.
CTAs typically trade on a wide variety of global exchanges and in a large number of markets. In the open outcry environment that dominated futures trading throughout the 1970s, 80s and 90s, it was extremely difficult and expensive for CTAs to even consider managing the idiosyncrasies and microstructures of all these exchanges and markets, as well as the associated risks. As a result, most relied heavily on their brokers and other intermediaries to execute their trades as well as assume the execution risk. If CTAs used technology at all for execution, it was to send a fax or e-mail a file of the day’s orders to their brokers.
After 2000, as electronic trading began to take root in the futures markets, some CTAs came to realize the benefits of screen trading. Among other things, electronic trading was significantly cheaper than phone-based or pit trading (pit brokerage and other order-handling fees could be reduced or eliminated), executions were much more transparent (No more front-running by locals!), quicker (No more “Where’s my fill?”) and post-trade processing was much cleaner. As a result, some CTAs became willing to assume more of the execution risk and began employing basic “point-and-click” execution technology, typically provided by ISVs and/or their brokers. But there was still a basic structural market problem that kept many CTAs from fully embracing self-directed electronic trading: Many key commodity-futures markets, including energies, metals, ags and softs, continued to trade via open outcry. This bipolar execution model gave many CTAs good reason to continue to rely on their brokers for execution.
Which brings us to the present. Through the past 12 to 15 months, just about all the major commodity-futures markets have successfully migrated the majority of their liquidity to the screen (See “By by floor,” above). For the first time CTAs can trade nearly 100% of their portfolio electronically. And in this new environment, the power of electronic trading, especially an algorithmic execution management system to facilitate that trading, can be fully exploited by CTAs. But before examining these benefits, it’s important to understand and underscore two key realities of electronically traded markets.
The first is that as markets become electronic, total volume typically increases, often dramatically, while the average transaction (ticket) size decreases. For example, average daily volume in the CME’s e-mini S&P 500 contract has surged from less than 160,000 in 2001 to 1.3 million contracts as of mid-2007. Yet according to a seminal study by Galen Burghardt, senior vice president and head of research at Calyon Financial in Chicago, the average trade size in the E-Mini is two contracts and the most frequent trade match is just one contract (see “Small doses,” below). The second reality is that much of this increased volume is created by totally automated trading systems and electronic market-makers employing algorithmic execution strategies. It’s therefore no coincidence that the recent volume explosions in energies, metals, grain and soft markets began shortly after the decision to “electrify” those markets.
Benefits of Algo Trading
So how can a CTA, who is not a high-velocity trader but will typically build and manage very large positions through time, benefit by utilizing algorithmic trading tools and execution strategies? Let’s start with the most important benefit: Improved trading performance. By employing some basic algorithmic execution strategies, a CTA can reduce market impact and reduce slippage, often significantly. For example, an algorithmic trading system makes it both possible from a technological standpoint and practical from an administrative standpoint to employ an execution strategy that breaks a very large order into very small pieces, effectively reducing the size of the order’s footprint in the marketplace. This capability is crucial in light of the aforementioned sharp reduction in average transaction size. But what if market conditions change and enough liquidity suddenly becomes available, within a predefined number of ticks, to complete the order immediately? The same algorithm can be programmed to identify and react in milliseconds to the emergence of the additional liquidity. Similarly, if the order is to buy/sell and prices begin to fall/rise, the strategy can be further enhanced to become more passive or aggressive. And if execution slippage can be reduced consistently, the door is then open for CTAs to develop effective higher-velocity, shorter-term trading models that wouldn’t be possible to execute through traditional execution methods or without an algorithmic execution capability.
Who develops these strategies? Many are provided by sell-side brokers as a means of providing value and securing clearing business; others come “out of the box” with the purchase or lease of an algorithmic execution management system. While a discussion about the development and implementation of specific algorithmic execution strategies could fill a book, it’s important to keep in mind that such strategies are designed to assist traders, not replace them. As such, the best strategies are generally those that can be customized and enhanced to meet each client’s specific requirements and allow the trader to react dynamically to changing market conditions.
A CTA who chooses to take direct control of the execution also assumes the inherent execution risk. This is no small consideration, especially in an electronic environment where the speed of execution is measured in milliseconds. Horror stories about “fat finger” errors and rogue algorithms sending out thousands of incorrect orders abound and should rightly give pause to the CTA contemplating self-execution.
In fact, error risk is perhaps the main reason CTAs have historically chosen to outsource execution to their brokers. While execution errors can never be 100% eliminated, an algorithmic trading system can provide powerful safeguards that significantly reduce the risk to a level that is tolerable vis-à-vis the reward of improved trading performance.
Let’s return to the original definition of algorithmic trading as simply rules-based trading. To the very large extent that risk management is predicated on a set of rules, algorithmic execution strategies can be written to include those rules. For instance, volume limits can be established by order, by market, by exchange, by trader, etc. Likewise, position and exposure limits (in any currency) can be established and monitored in real time. When these limits are reached, or even approached, video and/or audio prompts can be triggered requiring the trader to reconfirm the order, or the system can prevent the trader from executing the order. These “gatekeeper” rules can be extended to compliance. For example, an e-mail can be generated and delivered to a compliance officer if a risk-management rule has been triggered.
In a more holistic approach, the CTA may choose to integrate his black box (i.e., the computer that identifies the trade) and/or the order management system (the software platform that generates the order) with the algorithmic execution system. These connections are typically made via the FIX protocol or an application programming interface (API). For example, in such a set-up the black-box or order management system (OMS) could generate an order to buy 1,000 contracts of Nov. ‘07 WTI crude oil on the New York Mercantile Exchange with the order and all its parameters appearing directly on the trader’s desktop trade blotter.
In this instance, the execution-management software can be programmed to prevent the trader from any key-punch or point-and-click error. As such, the trader could not: sell any contracts; buy more than 1,000 contracts; trade a different contract month; and trade WTI crude on another exchange.
However, the trader could still retain complete control over how and when the order should be executed, perhaps using an execution algorithm to assist in the execution.
Of course, embedding a series of sometimes complex risk-management rules within an execution algorithm comes at the expense of speed. Every order must be examined and validated by the computer software before it will be released to the market. However, the loss of a few milliseconds (or less) is a small price to pay and typically inconsequential when compared to the risk-mitigation and peace of mind provided in return.
The benefits of algorithmic trading tools can be extended beyond execution and risk management to include the automation of many middle- and back-office functions, including account allocations and broker give-ups. Here again, the rules rule. Because allocation and give-up schedules are typically rules-based, orders that are executed algorithmically can be allocated to multiple accounts and given-up to multiple brokers in real-time. Not only can the time-savings be significant, but key-punching and other manual errors can also be eliminated. In addition, all the information on an electronic order ticket can be parsed, extracted and written to a file, which can then be used to create customized reports for delivery to other parties, including in-house or third-party back-office and fund-accounting systems. The delivery process itself can be further streamlined by integrating the algorithmic execution management system with the aforementioned front- and back-office systems via FIX or an API.
Now that nearly all futures markets are predominantly electronic, the execution environment has become conducive to a suite of customizable algorithmic tools that CTAs can adopt to help improve their trading performance, strengthen their risk-management practices and automate and streamline many of their middle and back-office functions. These rules-based tools are scalable and may be used independently or linked together and can be fully integrated into existing in-house and third-party applications. The CTAs who embrace this technology will enjoy a clear competitive advantage and be better positioned to react to the new opportunities brought about by electronic trading, as well as have more time to work on improving their trading systems.
Tom Griffo is vice president and head of global futures sales at Flextrade Systems Inc., in Great Neck, N.Y. Previously he was global head of marketing at Cargill Investor Services and has been involved in the futures industry for 25 years.
