The Gateway to Algorithmic and Automated Trading

The Five Drivers of Profitability

Published in Automated Trader Magazine Issue 12 Q1 2009

If speed isn’t everywhere, it’s nowhere. Phil Perkins, Ben Van Vliet and Andy Kumiega of the Institute for Market Technology argue that achieving a strategic alignment of trading, operations, and technology is a key factor in delivering competitive advantage.

Tech Workshop

When the Master governs, the people are hardly aware that he exists. - Tao Te Ching

The goal of trading is to make money. But how or why a trading system is making or losing money, why it has or has lost competitive advantage, may not always be well understood. It used to be that a person's trading acumen was the key driver of competitive advantage. No longer. With automated trading, many factors contribute to success, and the leader of a trading group must understand how not just trading, but also how mathematical algorithms, technology and operations contribute to the bottom line as well. Management has to keep the firm at the cutting edge of strategy, mathematics and technology just to stay alive. And they have to evolve as fast as the markets themselves.

Many firms evolve in reactive, ad-hoc management processes. These firms have no software design methodology, no long term goals, and are just barely keeping their head above water from day to day. These firms aren't aligning trading, operations, and technology. They operate like tornados.

Figure 1 Unaligned Organization

Trading, technology, operations, risk management, project management, and other functions randomly interact with each other in these organizations. The question is whether they are working with each other or against each other. With our colleagues at the Institute for Market Technology we have developed the I4MT Strategic Operations Framework (ISOF).
While techniques like Six Sigma address process variation, ISOF addresses the variation between what an organizations intends to accomplish and what it actually accomplishes. ISOF performs analysis of the gap between the stated business strategy and current working activities, because greater consistency in alignment will lead to greater consistency in profits from automated trading activities.

We summarize the drivers of profitability of systematic trading in 5 core goals.

We break down these goals in layers:

Layer 1: Goals. ISOF Goals in Figure 2.
Layer 2: Objectives. Each of the five ISOF goals breaks down into core objectives required to meet each goal.
Layer 3: Capabilities. Objectives break down into sets of organizational capabilities required to achieve the stated objective.
Layer 4: Activities. Capabilities break down into lists the process activities necessary to realize the organization capability.
Layer 5: Technology Architecture. Activities drive the development of the technology architecture necessary to most effectively execute the system.

By monitoring performance of activities and technologies, leaders can manage effectively and support the top level goal of increasing trading profits. Now, in this article we cannot derive all five layers, but will provide an overview of each of the five drivers in Layer 1. We recommend that you derive the sub-layers for your unique environment.

Strategic Information Advantage

Strategic information advantage means that your trading algorithms can be a source of competitive advantage. This includes research into quantitative methods, data cleaning, optimization, backtesting and risk management techniques. Essentially this is "how good are we at turning data into profitable knowledge?" This includes both hard data, like price data, and/or soft data like news items, sometimes called unstructured data.

As the advantage gained purely through technological superiority begins to wane, algorithms and backtesting will again move to the forefront. A priori research into trading algorithms can take on many forms: from reading academic articles, to deriving new equations and algorithms, to reverse engineering a competing system. Typically, research then moves to prototyping of models, followed again by attempt to improve results with new proprietary ideas.

Models then should be validated over an empirical validation stage, backtesting to ensure they really work. This should be true for news-driven trading strategies as well. A database of historical news in a standardized format can serve as a platform on which to backtest your text mining and trade decision algorithms. Given standardized news in one database and tick data in another, you can measure the magnitude of the impact of news on prices as well as the delay of that impact.

Some metrics to measure performance of your strategic information execution activities are:

• The ratio of the number of ideas to implemented models.
• Data usage metrics. Is the data we've paid for generating new ideas and profits?

Rapid Model Deployment

Rapid model deployment is another opportunity for competitive advantage. This is "how quickly do we turn trading ideas into working trading systems?" Trading opportunities come and go and good strategies don't work forever. The faster you get your system up and running the longer you'll have to capture your trading edge before the algorithm no longer works.

Rapid model deployment is accomplished through repeatable processes that allow for specialization of activities (Layer 4) and improved management oversight. Repeatable activity processes also lead to lower project and operational risks. Consistency is imperative if you are to measure and manage and improve your processes. Firms need assembly lines of trading models just to survive. The point is: use repeatable development processes to build repeatable trading strategies. Promote configuration over coding.
Some metrics to monitor the performance of deployment activities are:

• Proposal to Deployment Time. How long does it take for a formal proposal to become a working trading system?
• Intermediate Metrics. Understanding the steps along the process allows for assessment of individual activities.
• Operational Risk Metrics. The anecdotes here are too numerous to mention, but there is absolutely no excuse for losing millions of dollars because of a software bug.

Order Execution Speed

Of course, technological speed matters. This is especially true because the goal is not to beat the market, but to beat the competition. The first person, or in this case the first computer, to identify a trading opportunity generally doesn't leave much behind for whoever is second in line. This is true not only for deterministic strategies, like arbitrage, but also for well-known probabalistic strategies, like statistical arbitrage.

As Jeremy Johnson of the Tabb Group points out, for arbitrageurs "because speed determines the winners and losers, there will continue to be heavy investment in ... trading technologies. The firms that are capable of making the right investment and managing it through their infrastructure [will gain a] critical edge in their trading process."

The more deterministic the strategy, the more important technology becomes the determining factor of profitability. The goal should not necessarily be to be as fast as possible. Rather, the goal should be to be only fast enough to win. You may consider keeping one additional level of technology in your back pocket. That way if a competitor attacks your profitable system with their own, faster system, you have another level of technology ready to counter-attack with.

This is why so many firms often co-locate their servers with those of the exchange. Geographic distance between you and the exchange may be the largest source of latency. Logical distance-the number of network nodes-between you and the exchange is second. By removing geographical and logical distance, say through co-location and a direct exchange connection, you can remove latency. But, of course, this is expensive.
Latency causes slippage and slippage can be the difference between profits and losses. There is another tradeoff here, this time between technological speed and investment, but at least performance and latencies can more quantitatively be measured.

But, the real questions are these:

• How fast does my technology need to be to make money?
• How much will it cost to attain that speed?
• What will be the payoff if we do attain that speed?

If you can answer these questions, you will have the ability to understand whether a $10 million investment in the infrastructure for a direct connection is cheap or expensive.
We recommend defining technology quality attributes up front. A quality attribute is an observable characteristic. A quality attribute requirement is the threshold value a system must meet with respect to a quality attribute-functionality, capacity, extensibility, interoperability, scalability, speed, reliability, and so on.

Model Portfolio Management

No automated trading system exists in a vacuum and management has to decide which trading strategies warrant seed capital, because research and backtesting and data cost money. So, given many ideas, how do we effectively allocate and prioritize capital and time across the portfolio of new and working systems? In addition, given many working systems, the firm has to decide which ones will get investment capital and how much? This is exactly what "portfolio management" in this context means. It is the balancing act the maintenance of working trading systems versus the growth of new uncorrelated trading systems.

Diversifying a portfolio of trading systems, either by market, timeframe or strategy, can enhance the overall return on capital as well as reduce risk. Mapping allocations to strategic buckets can enable diversification, of course within the firm's core competencies. There's no use allocating money to a currency trading strategy if your firm consists of all option traders, though mapping could bring to light opportunities for new markets. Measures of performance in this area include: Average life of trading systems, profit per system, risk metrics (VaR), ramp up and ramp down times.

Minimize Transaction Costs

Minimization of costs is of course important. But, in the context of ISOF it may be more than a business aim, it may be a key component of competitive advantage.

A trading strategy may prove under backtesting to return 50 cents per contract traded. If transaction costs are 60 cents per round turn, this strategy will lose 10 cents on every contract. Clearly, if you can work your costs down to 40 cents, then the strategy suddenly becomes profitable. The key is to know through backtesting where that threshold is.
Then you can go about finding a cost structure that supports the strategy. The metrics for this driver of profitability are simply the quantified and allocated trading costs.

Implementation

Now, you may look at these four drivers of edge or competitive advantage and say to yourself "our firm is excellent in this area or that." Or you may say, "we need to improve in this area or that." Because most systems incorporate vendor supplied components, you should ask your vendors:

• How can your product or service increase trading profits?
• How do we best deal with the tradeoff of speed vs. context in soft data?
• How can we use your technology in our existing platform and how can we fit it into our development processes?
• How does your technology fit in with the five drivers of competitive advantage?

If you have a unified vision, then your vendors become plug and play.
Now, you may say that these four concepts are pretty abstract, how do we operationalize these drivers, given what we all understand the trading system development process to be: design, backtest, implement, manage. In our previous article, Rapid Model Deployment (RMD) (see Automated Trader Q4 2008), we stressed the importance of the K|V methodology for trading system development to design the trading strategy, backtest it to prove its repeatability, implement it in software on a hardware platform, and manage the risk of the working system. By aligning strategy and activities using ISOF, and having a well-structured, repeatable process like K|V in place, your firm can:

• Build trading systems that make more money.
• Shorten the trading system design and development cycle.
• Increase the speed with which new trading ideas are evaluated and either discarded or promoted, called strategy cycling.
• Lengthen the maturity stage of working trading systems.
• Increase the speed of recognizing and shutting down trading systems that no longer have a competitive advantage.

• Reduce the total cost of trading system development.
• Enable management to better manage the portfolio of trading systems
• Mitigate operational risks.

Alignment of strategy and activities, a repeatable process, and configurable, componentized technology framework will drive profitability. Finally, we recommend:

• Identifying opportunities that can drive competitive edge.
• Create your own strategic operations framework with layers that tie activities to the business goal of making more money.
• Operationalize goals through repeatable processes and componentized architectures.
• Monitor performance metrics.
• Root out causes of poor performance.

Doing these things should bring to light how each activity does or does not contribute to profitability.