You build a model specifying the factors that you think are related to returns, the groups or fractiles you want companies to be divided into, which companies you want to consider, and the historical context in which you want to test these factors. FactSet's Alpha Testing does the rest by calculating subsequent returns for each fractile, the entire universe, and the selected benchmark in easy-to-understand reports that help you identify factors correlated with excess returns, or "alphas."

Building a Model

Alpha Testing provides an intuitive interface through which you specify the factors you want to test, the historical context, and your customized fractile assignments.

When you build a model with Alpha Testing you designate each of the following:

• Time frame: Indicate the start and end date during which your model will be run, as well as the return and rebalance frequency. These specifications also include how many days data-availability will be lagged in the test to eliminate look-ahead bias.
• Universe and benchmark: Decide which companies the model will evaluate and which benchmark will help you to best measure the success of the idea. Perhaps your model considers a universe of only German companies or perhaps only the 600 largest American companies by market capitalization. Whatever your specifications, Alpha Testing has the flexibility to handle them, along with the appropriate benchmark.
• Ranking factors and group divisions: Designate which factors will be used to group the universe constituents, as well as the number of groups you want to consider. Alpha Testing is flexible enough to handle both single and multi-factor models. Alpha Testing generates a report for each group and its returns.
• Data implementation: Choose which research databases (such as Compustat, Worldscope, and MSCI) are queried. The wealth of data you can choose from nearly eliminates any potential historical biases.

Analyzing the Results

Further your analysis with detailed period reports for each fractile and detailed fractile reports for each period. If you chose two ranking formulas, you can also view fractile-by-fractile reports. In every report, you can see the underlying constituents, change report sorting, or adjust the weighting factor of the returns.

Alpha Testing reports include CPU-intensive mathematical and statistical calculations. You can view return information in either overview charts and reports or in detailed reports for any specific fractile or time period. You can even see the underlying company constituents of any fractile at any time.