Gerontocrat,
Plotting by year makes the trend harder to see or measure. Try plotting the minima on the x axis and the ensuing maxima on the y. Using all years, the trend is small. However, starting with the years with low minima (2002), the data points are not clustered equally throughout the four quadrants. Compared to the trend, 7 years exhibited a lower minima and higher ensuing maxima, 5 showed a higher minima and lower maxima, 4 showed both a lower minima and maxima, and 2 both higher minima and maxima. Twice as many data points appeared in the quadrants with opposite signs as those with similar signs. Additionally, of the 9 years with a higher maxima, the 7 highest maxima where all preceded by lower minima. The lower the minimum, the more likely that the ensuing maximum is higher.