Hello,
For example, in the 'simple exponential smoothing' the formula is: L(t) = alpha*Y(t) + (1-alpha)L(t-1)
But, you need, an initial L(0) to construct the next L(1).
In Double, Holt, Winters, etc...you always should have a rule to decide the first element, isn't it?
In the following text they call it 'Initialitzation': https://www.otexts.org/fpp/7/1
Regards,
Ferran