Choosing What to Trust: Why Assembly History Has to Lead
TL;DR
Once every election speaks the same language, should all three results be blended equally? Leaning on recent momentum makes the estimate jumpy. Anchoring in Assembly history provides something stable.
The first instinct is straightforward: recent elections should carry more weight.
When recency feels noisy
That approach seems reasonable. Recent results feel relevant. Assembly history feels old. But experimentation shows that relying heavily on recent results makes the estimate jumpy. Small changes push the numbers around more than back-tests show is stable.
A different framing
Instead of asking which result is most recent, the better question is: which result is structurally closest to what is being estimated? Assembly elections define Assembly outcomes. Everything else adjusts around that core.
So the blend flips. Assembly history becomes the anchor. Local and Lok Sabha results become refinements, not replacements. The estimate stops overreacting.
What this provides
The result is a grounded starting point: a blended Assembly-equivalent base. Recent results integrate into the last Assembly result without losing structure. But this is only the starting line. Turnout, long-term drift, and campaigns have not yet entered the picture.