Monday, August 15, 2011
A Perfect Time to Stress-Test Strategy
TradeStation integrates a powerful new technique for testing your trading theories in real markets.
If polishing up your theoretical trading models, rather than trading, was the most profitable use of your time in the last few weeks, TradeStation has just the product for you.
The sophisticated online broker last week integrated what it calls a Walk Forward Optimizer into its trading platform. Walk-forward analysis is a powerful technique that allows a trader to see how well a forecasting model works, by comparing its projections to historical data in a very precise manner.
TradeStation (http://www.tradestation.com), winner of Barron’s 2011 Online Broker Survey, acquired South Africa’s Technovest in June 2010, and hired its CEO, Wouter Oosthuizen, to bring the Internet-based financial technology he’d developed onto its latest downloadable platform, TradeStation 9.0. The Walk Forward Optimizer, or WFO, is designed to tell a trader whether his strategy will be reliable in practice. It automates a model-testing process that would be incredibly time-consuming if done manually. The Electronic Investor was given a late-beta demonstration of the tool courtesy of Oosthuizen, who is now a senior software engineer at TradeStation.
How does the WFO work? Let’s say you have 12 months of data on which you’d like to test your strategy. You might first focus solely on how your system works over the first four months of historical data, tweaking inputs, or parameters, to get the best possible result. These inputs might include anything, from a sharp move up, to insider transactions. You’d then “walk forward” to include a new month of data and see how your model works under the new conditions. Eventually, you’d progress through the whole 12 months, tweaking as you go. An actual walk-forward test would involve a lot more data, and multiple tests over the same set to get your trading model right.
The analysis calculates five sets of pass/fail evaluation criteria: overall profitability; correlation between in-sample (the development of your model) and out-of-sample (the actual test) data, which is also known as walk-forward efficiency; consistency of profits; distribution of profits; and maximum drawdown. Prior to running a WFO, the user decides on what the various thresholds are for passing or failing. Once that’s done, the WFO runs on multiple sets of data. It all gets done surprisingly quickly. Obviously, a “failed” reading means that strategy isn’t very robust and a “pass” means that it is. You can look at 3D graphs to help you analyze the results.
TradeStation has a video explaining this new feature on its QuickTips page.
The system can also do sensitivity analysis, which considers the effect of an individual parameter, say a moving-average crossover, on the model. By varying the weighting of a parameter, a trader can identify individual problems and evaluate the profitability of the model under different parameter weightings. It also provides distribution analysis that lets you display a variety of graphs comparing specific performance values against underlying market criteria. You can, for example, look at trades placed at specific times of day to see when the strategy is most successful.
This process can be highly technical; I didn’t run across most of the particulars until my second year of graduate work in econometrics. But you don’t have to understand all the nuts and bolts to be able to use this tool.
Oosthuizen says, “We don’t know of any other broker that has walk-forward analysis incorporated into its platforms. We believe the TradeStation Walk Forward Optimizer is a real game-changer for strategy traders, offering the ultimate strategy-trading stress-testing tool.”
MARKETSMITH GOES MOBILE. The company, which produces an incredibly flexible stock screening and charting tool, gave Barron’s an exclusive look at its new iPhone and iPad apps. We reviewed MarketSmith’s browser-based service, which runs $999 annually, earlier this year ("Two New Ways to Make Money,” April 11, 2011). Subscribers can use any of these platforms for no additional fee.
We like the iPad app, which is best viewed in landscape mode; the iPhone version is just too cramped for our aging eyes. MarketSmith (http://www.marketsmith.com) is a graphics-intensive approach to analyzing stocks and mutual funds, blending technical and fundamental stock data as well as proprietary ratings and rankings. The graphs look great on the iPad.
Subscribers can search the universe of stocks and flag those of interest on the go, then review the list in more detail on their desktop. This is a “light” version of the desktop application, but it allows you to perform screens and view complex charts when you’re mobile, and perform more complicated analysis once you’re back at your home base.
The firm’s president, Scott O’Neill, notes that “when a market gaps up or, as mostly recently, gaps down, a mobile application that helps keep investors grounded in the facts is invaluable to their profitability and peace of mind.”
You can access a two-week trial version of MarketSmith, including the mobile apps, for $19.95. The trial offer also includes product coaching so you can learn how it works before paying for the full year. If you want to check it out without laying out any money, there is a free three-day trial available as well.
GETTING OUT BEFORE THE FALL. SmartStops (http://www.smartstops.net), a service that calculates stop-loss targets and generates signals for re-entering a position, just launched a feature called the Market Risk Barometer. It measures the risk posed by components of an index or industry against its recent risk history. It’s calculated by dividing the current percentage in an above-average risk state to the 100-day moving average for that industry or index. A value of 1 or higher indicates a risk higher than the 100-day moving average.
Not surprisingly, every industry and index tracked by the new barometer was well above 1.0 when we reviewed the new tool on Aug. 8. It was alarming. Ninety-seven percent of the Dow Industrials were considered “above average risk,” while the barometer measured a whopping 1.81. Ouch.
A WILD WEEK: Online brokerage giant TD Ameritrade reported some notable occurrences during the 635-point drop in the Dow on Aug. 8. The firm’s trading volume exceeded the records set in the flash crash; previous records were shattered for options and futures, and mobile usage surged 29% from the previous week.
Published in Barron’s Online, August 11, 2011