A backtest encompasses
an entire user-defined universe of stocks. The universe of stocks for each
backtest can be built using SYFIR's intricate universe selection module
or by setting the universe of stocks to a standard benchmark or benchmark
subset like the S&P500 or the Russell 2000 Value. The backtesting universe
of stocks can then be compartmentalized into subsets or "models" as defined
by SYFIR's manager-specified peer group selection system. This methodology
allows different data modeling strategies to be employed for each of the
individual subsets of stocks. For example, a manager performing a backtest
on the S&P 500 may want to apply a different strategy for the Utilities
sector than he would for the Financials Sector. Each subset of stocks is
then modeled against its peers with their resultant scores normalized and
merged together to represent the whole composition of the backtest.
The diagram below shows an example of the universe selection, factor selection, and factor weights for a backtest containing two models. The large box labeled Backtest: Multisector S&P 500 represents the bactest itself and illustrates the one universe selction criteria that will be applied to all models in the backtest (S&P 500 Index Membership). The box on the lower left labeled Model 1 contains a model labeled S&P 500 Financials 1990 - Present. This box illustrates the users universe selection criteria, factor selection criteria, and weights for the financial Sector model which will make up a portion of the users backtest.
Multisector S&P 500
The two boxes to the right contained in the larger box labeled Model 2 illustrate the universe selection criteria, factor selection criteria, and weights for the utilities sector model which will make up another portion of the users backtest. The box labeled S&P 500 Utilities 1990-1995 represents the users selection criteria, factor selection criteria, and weights for the years 1990-1995. The box directly to the right labeled S&P 500 Utilities 1996-Present illustrates how the user is able to change factor selection criteria, and weights for the utilities sector model for different time periods. SYFIR Backtesting will overlay the two date periods to form one model which covers the entire selected date range for that sector model and will be used as one model in the constitution of the overall backtest.
Custom House St.
Providence, RI 02903
Fax (401) 831-4376