Data Limitations and Biases

The are two years of data available for the backtest. The backtest calculates the Total Return as % Price Change + Yield * ( 52 / Holding Period)in weeks. Since we are using 1,4, 12, or 24 week holding periods, the above Total Return is only an approximation, but we think it is very close to the actual total return. The database used for Backtest has not been adjusted for splits. However, the backtest results are calculated using the Percent Change in Price item which is split adjusted. The database used for Backtest contains only survivor companies - it is created using the DBCM universe, so all backtests using it have survivor bias. On the plus side, the database used for Backtest is the only estimate database in existence that does not have a look ahead bias. For example, in other databases, the Current Fiscal Year (F1) estimates on a specific date t include estimates that were received on dates t +1 , t +2, etc., that were dated before t by the broker but which had not yet been received on date t. The (F1) estimates in the backtest database are what was known on date t.

---

Back | Backtesting | Help Home Page