Tamil Nadu’s 2026 Assembly Election was the largest single-cycle electoral reset the state has produced in modern times. 164 of 234 seats flipped party. A first-time party won 108 of them. The sitting Chief Minister lost his own seat by 8,795 votes. The new Assembly’s median MLA age fell by twelve years.

This page is the canonical hub for everything published on this site about that election. One interactive dashboard. Five investigations. The open dataset behind all of it. Plus the pre-poll work that shaped what to look for once the results came in.

If you arrived here from search, social, or a citation, start with the dashboard. If you arrived here from one of the individual posts, the other four investigations linked below cover the angles that piece did not.

Interactive · 234 seats · 12 parties · 4.93 crore votes
Open the TN 2026 Explorer
Seven sections in one page. Constituency map, party seat-vs-vote efficiency, 9-region cross-party anatomy, 2021 to 2026 swing per seat, demographics of all 234 winners, defeated incumbents, dummy near-misses. Hover any seat, sort any table, filter any view.
Open the dashboard

The findings, in five investigations

Each card targets a distinct angle of the same election: its headline number and the one finding it turns on. They cross-link tightly, and the full posts carry the regional and demographic anatomy behind each figure.

108of 233 seats won TVK 233: A Debut, Mapped A tsunami in Chennai, a ripple in the Cauvery delta. The debut wave mapped to exactly where it broke, at a 1.32 seat-to-vote efficiency. Read the analysis → 133 → 59seats, in one cycle How DMK Lost Tamil Nadu, Region by Region The taller they stood in 2021, the harder they fell (a +0.67 correlation). Chennai dropped from 31 wins to 2. Read the analysis → 56 / 234incumbents kept their seat The Chief Minister Lost His Seat M.K. Stalin lost Kolathur by 8,795 votes. Of the 93 incumbents who fell, 64 went straight to TVK. Read the analysis → 52median MLA age, down 12 years The Youngest, Most-Educated Assembly TN Has Elected 41 MLAs under 40 and 22 women elected. A generational reset in a single election. Read the analysis → 0 / 2seats flipped (strict / loose) How Many of the 329 Dummies Actually Mattered The mechanism stayed intact, the wave drowned it. Zero consequential seats by the strict test, two by the loose one. Read the analysis →

The dataset

The full 4,257-row candidate dataset plus the 329 flagged name-match pairs are released under CC-BY-4.0 on GitHub. Free to use, cite, and remix.

Two CSVs. Per-candidate metadata scraped from the Election Commission of India’s Affidavit Portal, plus the name-similarity match pairs that drove the dummy-candidates investigation. Sourced cleanly. Schema documented. Reconciled against the published results within 1 to 2 rows on every headline figure.

Open the dataset release page →

Repository: github.com/ndranandraj/tn-2026-candidates-dataset

Methodology, briefly

Three technical pieces hold the analysis together.

A custom nine-region political geography for Tamil Nadu (Chennai and Suburbs, North, Central, Cauvery Delta, Krishnagiri Belt, Northeast Coast, Madurai Region, Kongu, Deep South). Standard published groupings stop at five or six regions and obscure the most interesting cross-regional asymmetries. The nine-region cut is what makes the DMK collapse visible as a non-uniform phenomenon rather than a flat statewide drop.

A three-tier dummy candidate classifier (EXACT, NEAR_FULL, WORD_MATCH), each tier using a different name-normalisation pass. The strict test (EXACT plus NEAR_FULL only) is what supports any causal claim. The loose test (all tiers) is what catches the noisiest signal.

Token-based incumbent matching across 2021 and 2026 ECI filings. Indian politician names drift across cycles (initials swapped, caste suffixes added or dropped, transliteration variants, post-marriage name changes). Exact match misses roughly half the real cases. Token-set fuzzy matching scoped to “same person ran in some constituency in cycle N and again in cycle N+1” recovers the rest. The published incumbents post reconciles within 1 to 2 rows on every headline figure.

A separate methodology post walking through the incumbent-matching algorithm is in the queue. When it ships, it links here.

Pre-poll companion pieces

Three posts published before the May 4 vote, mostly to set up what to watch for. They read differently from the post-results work because they were written without knowing the answer. All three predictions land partially: TVK did spread its support widely, the dummy mechanism did not bite, and the regional asymmetry in DMK’s footprint did predict where the collapse hit hardest.

A note on what this analysis is and is not

This is a data-driven account of what happened. It is not a political analysis of why it happened. The structural drivers (anti-incumbency, alliance fatigue, Vijay’s brand reach, cadre defections, the AIADMK organisational drift) are layered and seat-specific, and serious causal claims require more than vote tallies.

What the data does support is the regional and demographic anatomy: where the wave concentrated, which incumbents could not survive it, what the new bench looks like, and what the dummy-candidate mechanism did and did not do under wave conditions.

If you use any of this work in your own writing or research, citation back to this site is appreciated. If you find an error, the dataset repository is the right place to file an issue.