SIDC monthly SSN is filtered with a triple running mean filter to remove variation faster than 12 months.
The Hathaway estimation for remainder of cycle 24 is appended and cycle 24 is repeated as a simple estimation for cycle 25, as detailed below
This permits an good 11y filter (triple running mean) to be compared to a straight running mean.
The artefacts of the running mean are clear and in many places are out of phase with the SSN data: eg. the 1986 solar minimum and 1990 max. Indeed, it is clear that there is variability on a scale of less than 11y in this part of running mean output.
The relatively steep drop from 2003-2008 can be seen to be an artefact of the defective filter.
Extending using estimated SSN:
The Hathaway estimation of the remainder of c24 stops in 2020. The SSN at that point is drawn back to where it meets the actual data of c23: 2007.75
2020-2007.75 is taken as the length of c24.
2008.8 is minimum of the c24 estimation.
The simplistic assumption that c25 will resemble c24 is used to collate a replica of c24 to the Hathaway estimation.
This gives a rough estimation of magnitude and timing of c25 that helps fill out filter buffer and since it is just the tail, it should not perturb the result unduly.