The period from 1970-90 showed the highest correlation to Nino 3.4. The graph above is an example of a test of forecasting 250 mb U anomaly for the 1958 - 1998 dry seasons using just the equations from just 20 years of data between 1970 and 1990. In other words forecasting backward and into the future. Notice the results are very accurate for the 70-90 period, and noticeably worse before 1970 and after 1990.  Using 42 years of data for the forecasts give more reliable results, however, there is no guaranty that there are not other physical cycles that may change the accuracy of the equations in the future.