Finding — Vote efficiency: who converted votes into seats, and who didn't
TVK had a 46.4% win rate (108 of 233 contested). DMK 33.5%. AIADMK 27.3%. BJP 3.0% (1 of 33). Naam Tamilar Katchi contested all 234 — won zero. And someone got 42.5% of votes in their AC and still LOST.
Win rates by party
| Party | Contested | Won | Win rate | Median % when WON | Median % when LOST |
|---|---|---|---|---|---|
| IUML | 2 | 2 | 100% | 33.9 | — |
| TVK | 233 | 108 | 46.4% | 40.5 | 29.8 |
| CPI | 5 | 2 | 40.0% | 38.4 | 29.4 |
| DMK | 176 | 59 | 33.5% | 36.2 | 30.0 |
| AIADMK | 172 | 47 | 27.3% | 36.3 | 27.0 |
| VCK | 8 | 2 | 25.0% | 39.0 | 31.0 |
| PMK | 18 | 4 | 22.2% | 37.5 | 26.6 |
| INC | 28 | 5 | 17.9% | 36.1 | 26.7 |
| BJP | 33 | 1 | 3.0% | 32.6 (only 1 win) | 21.8 |
| NTK | 234 | 0 | 0.0% | — | 3.5 |
What this says
TVK was the most efficient vote-to-seat converter
46% of TVK's contested ACs flipped to TVK wins. Vijay's first election achieved a win rate higher than DMK or AIADMK despite running solo without any alliance arithmetic. Compare to BJP (3%) and NTK (0%) — both contested widely and converted little.
The "got 42% and still lost" phenomenon
In a multi-party first-past-the-post system with no alliance arithmetic, you can win with 34% or lose with 42%. The data:
- Maximum vote share when LOSING: TVK 42.5%, DMK 40.3%, AIADMK 35.8%. Specific ACs where someone got >40% and still didn't win.
- Minimum vote share when WINNING: TVK 29.3%, DMK 29.4%, AIADMK 31.4%. Some winners squeaked through under 30%.
This means: in a tight three-way TN race, the difference between winning and losing is often 5 percentage points. The vote-to-seat function is steeply non-linear around the 33-35% mark.
NTK — contested all 234, won zero
Naam Tamilar Katchi was on the ballot in all 234 ACs. They won zero. Their median vote share: 3.49%. Maximum: 13.4%.
Either you're a swing-mode third-party draining votes (and NTK does that — they routinely poll 3-5% statewide), or you commit to a focused contest to win specific seats. NTK's universal-contest strategy got them statewide presence but no representation. They're not a party; they're a protest vote channel.
BJP's collapse (in seats won)
BJP contested 33 seats in 2026 (up from 20 in 2021 as part of the expanded NDA allocation). They won 1. Their median vote share where they contested was 22% — well below the 29-33% range of winning candidates.
For comparison:
- 2021: BJP contested 20, won 4. Win rate 20%.
- 2026: BJP contested 33, won 1. Win rate 3%.
The contesting-more-seats strategy backfired — BJP traded a 4-seat caucus (with 20% conversion) for 1 seat (3% conversion). The "more seats means more wins" calculus failed because TVK absorbed the anti-DMK vote that had previously gone to BJP candidates in NDA fights.
CPI(M) + VCK — the DMK alliance allies
CPI and CPI(M) are part of the DMK alliance. CPI got 5 seats with 40% conversion (2 wins). CPI(M) won 4 seats (per total). VCK won 2 of 8 (25%). The Left/SC allies held up better than DMK itself in conversion rate, suggesting they ran in safer DMK-stronghold pockets that even the TVK wave couldn't crack.
The compactness angle
How tightly clustered is each party's vote share across ACs?
| Party | Median % | P10–P90 spread | Standard deviation |
|---|---|---|---|
| INC | 28.2 | 10.5 | 4.6 |
| DMK | 31.8 | 11.7 | 5.1 |
| BJP | 22.0 | 15.3 | 6.1 |
| TVK | 33.7 | 19.4 | 7.5 |
| AIADMK | 28.8 | 21.0 | 8.3 |
| PMK | 28.2 | 16.7 | 9.0 |
| NTK | 3.5 | 4.0 | 1.7 |
- DMK and INC are the most compact — they ran similar vote shares across many ACs. Hallmark of a disciplined statewide machine.
- AIADMK is the most variable — wide vote share spread (P10 to P90 of 21 percentage points). Pockets of strength and weakness.
- TVK is in between — broader spread than DMK but not as wide as AIADMK. Their wave reached most places but unevenly.
- NTK is hyper-compact at low share — same 3-4% everywhere, no concentration.
What you can't see in headline vote-share numbers
The win-rate × median-share-when-winning matrix is the cleaner story than "TVK got 34%." It tells you that TVK converted at 46% with a 40.5% winning-pace — much higher than the 33% statewide median, because they won the contested races by larger margins than they lost the unwinnable ones.
Cross-check
All numbers from s3://tnelection2026/results/curated/year=2026/ac=*/candidate_totals.parquet joined to per-AC winners (rank-1 by total_votes). pipelines/deep_dive.py.