The use of mathematical models for constructing actual trades is the core business model of quantitatively focused trading firms, but this doesn’t preclude humans from playing a critical role. Because the strategies are statistical in nature, they require a great deal of quantitative analysis and back-testing by developers.

“We do not treat our systematic trading strategies as ‘black boxes’,” said Richard Flom, vice president and head of trading at Systematic Alpha Management, a New York-based systematic Commodity Trading Advisor. “An analogy I use is a pilot flying a plane. The plane can be flown on autopilot but you’d never be able to rely on that 100%, so you want to make sure the pilot is always in control. You always need a pilot for takeoff and landing as well as any issues that may arise during flight.”

Systematic Alpha Management monitors all trading throughout the day to verify that all executed trades are performing as close as possible to predicted results. “One of the tasks of trading is to make sure that trading closely matches what back-testing is going to show later,” Flom said. “Trading should be as close as possible to hypothetical and actual research.”

The Systematic Alpha Futures Fund, Ltd., embeds the company’s core trading strategy, which has existed since 2001. This strategy is focused on trading relationships between most liquid futures markets via short-term mean-reversion strategies. Investors looking to blend a short-term mean-reversion approach with a longer-term directional or trend-following systematic trading can invest in another fund, Systematic Alpha Multi Strategy Fund, Ltd. Managed account clients can choose the strategy and leverage level.

The categories of liquid futures that the company trades include equity indices, currencies, commodities and fixed income products. It trades triangular relationships in our mean reversion spread strategies, for example the S&P E-minis versus FTSE-100 and British Pound.

The company periodically reviews new futures contracts to see if any have increased in liquidity and have enough historical data for it to back-test and possibly trade. “Theoretically, it is possible to have a new liquid futures market, but unless we have backtesting and enough data for it (5-10 years), we will not trade it,” said Flom.

Developing viable systematic strategies, Flom said, includes such steps such as hypothesis proposal and testing, implementation of trading strategy(s) to exploit, back-testing, transaction costs modeling, test-trading of the prototype and actual trading with tracking of the transaction costs.

“Once enough actual data for transaction costs is accumulated, one should make sure that the assumption of certain transaction costs and the actual experienced transaction costs for a given level of assets match,” Flom said. “This makes the quant model self-consistent. Being able to weed out those models that we should implement in trading, and picking the optimal parameters for the models that have good results in back-testing over the years is some of what research does here.”

Although the research team is not sitting on the trading desk, it’s constantly looking at various performance measures and back-testing results, picking production parameters, looking at possible new relationships for trading, as well as improvements to the systems. Developers, who are also not sitting on the trading desk, are developing software solutions based on input coming from both research and trading. All of its systems, with the exception of the front end trading platform, are developed in-house.

The models are dynamic in nature, in the sense that, not only are they learning market conditions as trading continues from open to close, but they’re subject to retesting and periodic adjustments. “Depending on the models and their performance, every time we get new data, we re-back-test our parameters, and tweak them a little bit if necessary to find the most optimal equity curve for actual trading,” Flom said.