Legal Beagle by Graeme Edgeler

I've done quite a bit of finely-grained Booth-by-Booth analysis over recent years

It's allowed me to identify various Bellwether neighbourhoods

Who knows - Intensive canvassing of these micro-areas might just prove a little more cost-effective for Labour & Greens than Polling

...the “wasted vote” (ie votes for parties that don’t cross the 5% threshold and don’t win an electorate).

I have to admit I hadn't fully realised that this was part of the MMP equation - until they started talking about the 'discarded vote' on Morning Report this morning.
Though I thought it was a bit on the nose at this point in the voting process for them to be pointing to a Green vote as a potential 'discarded vote' - that may sway some voters.

and now I know that bellwether refers to a sheep not an atmospheric state

There is a probability that an electorate that may be thought a "bellwether" has actually aligned that way from chance (more so with FPP electorates that always elect an MP from the winning party, I think).

It would be interesting to calculate this for Ōtaki and also to look at the time series.

I sort of get your bellwether electorate argument. If I take Mangere at the other end – it says nothing about votes so much as it is more likely to have voted Labour (in its case) irrespective of what the national party vote was.
If that's the case then does your model give us a better sense of what the swing nationally could mean or is that casting too long a bow?

If that’s the case then does your model give us a better sense of what the swing nationally could mean or is that casting too long a bow?

I haven't gone nearly that far. Simple math, not full analysis :-)