The presented research results indicate that for fiancial time series two years of training data is frequently all that is required to produce optimal forecasting accuracy. Results from the Swiss franc models alert the neural network researcher that the TS Recency Effect may extend beyond two years. A generalized methodology is developed for determining the minimum training set size that produces the best forecasting perfornance. The generalized method was tested on two stock indexes (DJIA and CAC-40) and an individual stock value (DIS) to demonstrate the functionality and practicality of the methodology