What the report does

Three views on the same evidence: candidates running ahead of their own party locally, candidates running behind it, and councils where a party has stayed above its comparator group for several elections in a row.

How it's built

1. 1

   Carry the scores across

   Scores and runs come straight from Bucking the Trend. Nothing is recalculated here, so the two reports can never disagree.

2. 2

   Work out length of service

   Count the elections each person has stood at. Where the span is implausible, the contests don't hang together geographically, or a contest is clearly theirs but unlinked, the figure is marked uncertain rather than dropped.

3. 3

   Sort into signals

   Candidates running well ahead of their party locally become 'talent to watch'; those running well behind become 'worth a conversation'. Anyone who fits neither is left uncategorised on purpose.

4. 4

   Read the councils separately

   Council rows use the council's own run of above-trend elections. They carry no identity dependency, so nothing is held back there.

5. 5

   Hold back what is visibly wrong

   Where a known identity problem makes a number plainly wrong rather than merely uncertain, the record is not shown at all, and the count of those is published.

How to read it (and how not to)

• • Most movement in a local election is national, not local. This surfaces exceptions; it does not claim every positive score is hidden talent.
• • 'Worth a conversation' is information, never a recommendation about anybody's future. A gap can come from a strong opponent, a local issue or an unusual field.
• • Length of service is shown even when uncertain, marked as such with the reason. It is never silently dropped.
• • A small number of records are hidden because a known identity problem makes the figure visibly wrong; the count is published rather than buried.
• • Northern Ireland is excluded, as in every other report here.