Model output · compiled 25 Aug 2026 · not a forecast of fact
Lok Sabha 2029 — seat-by-seat projection
Every one of the 543 constituencies is called bottom-up on its own evidence: its real 2024 ECI margin, the actual assembly-segment votes cast in that seat since (Maharashtra, Haryana, Jharkhand, J&K 2024 · Delhi, Bihar 2025 · Bengal, Assam, Kerala, Tamil Nadu 2026), documented local factors and 2025–26 events. The Lok Sabha totals are simply the sum of those 543 calls — aggregation, not allocation. Tap any seat for its full signal table.
The projected Lok Sabha
543 seats. The dashed line marks 272, the majority threshold.
338NDA
275BJP alone
166INDIA
19TVK
20Unaligned
Model-computed Lok Sabha. 543 seats, each called on its own evidence — its real 2024 result, its own assembly segments where a state has voted since, documented local factors and 2025–26 events — then summed, with no national-mood constant of any kind. The BJP lands at 275, three past a solo majority, with NDA allies at 63. Toss-up seats decide whether that solo line actually holds — the alliance clears 272 either way.
Probabilities, not a point call. 20,000 correlated simulations (national ±3 · state ±3.5 · seat ±6 — variance, not bias, since 2019 and 2024 falsify any directional national constant): P(NDA ≥ 272) = 93% · P(BJP alone ≥ 272) = 56%. 80% bands — NDA 281–378 · BJP 233–318 · INDIA 132–229. The 338 / 275 headline is a median, and the solo-majority question is a genuine coin flip.
Alliance scenarios — the historical forecast-killer.(All scenario arithmetic is computed off the audited base; reset the simulator to compare like-for-like.) Every major forecasting failure (2004, 2015 Bihar, 2021 Bengal, 2024) came from state systems and alliance arithmetic, not national mood — 2024 itself was a −0.8-pt national swing hiding a 63-seat change that decomposes almost entirely into ten state-cluster stories. Stress-tested here: JDU exits NDA → NDA floor 327, likely ~332 (BJP retook most of this belt contesting without JDU in 2014). ADMK and BJP contest Tamil Nadu separately → TVK 19→24, DMK-bloc +4, ADMK 8→2, NDA −6. UP-2027 repeats UP-2024 → NDA 43→35 inside UP alone. And the priors expire on schedule: UP · Punjab · Uttarakhand vote in 2027, Rajasthan · MP · Chhattisgarh · Telangana · Karnataka in 2028 — every modelled swing gets replaced by measured segments before 2029, and the model re-solves on each.
Swing simulator
live — recomputed from the same 543 seat records
Base case = the audited board above. Drag the swing, or edit the bloc shares — margins move seat by seat, flips are listed, totals and probabilities recompute. Nothing here is a separate model.
National vote split (editable — always sums to 100)
Bloc
Base %
Simulated %
NDA
43.3
INDIA
41.6
TVK
2.3
Others
12.8
12.8auto-balance
Total
100.0
100.0
Base = ECI 2024 national shares by alliance, with TVK's 2026-implied entry (~34% of TN ≈ 2.3 national pts, drawn ≈ 40 / 35 / 25 from INDIA / NDA / others). Swing here means share transferred across the NDA–INDIA axis: +1 to NDA = −1 to INDIA = +2 pts of margin in every seat contested on that axis. TVK edits convert ×543/39 ≈ 14 onto TN margins. TVK and unaligned seats are untouched by the national slider — their contests are not on that axis.
Party-wise base vote % the model anchors on (ECI 2024 · TVK 2026-implied)
BJP 36.6
INC 21.2
SP 4.6
TMC 4.4
YSRCP 2.6
TVK ~2.3
TDP 2.0
DMK 1.8
CPM 1.8
RJD 1.6
BJD 1.4
JDU 1.2
AAP 1.1
UBT 1.0
NSP 0.8
ADMK 0.8
SS 0.7
CPI 0.5
JMM 0.3
Others ~11.5
The model never assumes a national poll number — it works in seat-margin space from these 2024 anchors plus measured segments. This table is the translation, shown for coherence; edits happen at bloc level because intra-alliance transfers do not move the seat axis.
Map explorer
Every one of the 543 constituencies is drawn in its real location and coloured by projected winner. Start on the all-India map — tap or hover any constituency for its call and reasoning; open a state chip for that state — on its actual polygon or on the tiles — for the full signal table and the logic behind that call. Chips cover every state and UT.
How every seat is scored. Projected margin = 0.55 × (real 2024 LS margin, signed for the pick) + 0.45 × (the seat’s own post-2024 assembly-segment lead, discounted ρ = 0.6 for 2024 assemblies, 0.7 for 2025, 0.75 for 2026) + no national-mood constant — 2019 rose after five years in office and 2024 held outside specific state coalitions, so a mechanical incumbency term is not evidence; national drift since 2024 enters only where it is measured (the segment layer) or as labelled state-specific modelled swings + documented local factors + documented 2025–26 events (post-Nitish JDU risk, Ladakh consolidation). Segment leads are computed from AC-level ECI votes aggregated through the official DataMeet AC→PC map (4,182 ACs); Haryana and Jharkhand use a win-count proxy calibrated on 86 PCs (lead ≈ 27.2×ΔW/n, R² 0.33); Assam is exact for 3 PCs and district-approximated for 11; Seemanchal and Kokrajhar segments are half-weighted for third-force distortion (AIMIM, BPF-vs-UPPL). Tamil Nadu and Puducherry are scored directly on 2026 three-bloc shares — the 2024 base is dead there. States with no election since June 2024 (UP, RJ, MP, GJ, KA, TG, AP, OD…) keep the modelled state-swing layer, labelled as such on the seat card. Seats are flipped wherever the evidence says so and re-solved to a zero-contradiction fixpoint; totals are pure sums of the 543 calls. Why the 2024 base anchors the blend: same-office persistence is the strongest single seat-level predictor, and assembly voting demonstrably diverges from Lok Sabha voting (Bengal is the live proof — BJP won the 2026 state, yet Bengal’s own LS record argues caution), hence the ρ discounts. The local and events layers are deliberately restricted to documented, citable seat-specific facts — sparse by discipline, not neglect. Confidence: solid ≥ 10, lean 4–10, tossup < 4. Uncertainty is carried by 20,000 correlated simulations and explicit alliance scenarios, because the failure record of Indian forecasting — 2004, 2015 Bihar, 2021 Bengal, 2024 — lives in state systems and alliance arithmetic, never in a national mood number.
Rendering constituency geography…
Who is Prime Minister after the 2029 election
Seventeen named individuals, single numbers, summing to exactly 100. Conditioned on 20,000 simulations of the audited board (live-linked to the simulator on the Lok Sabha tab): NDA-led government ~94%, INDIA-led ~4%, hung or other ~2%. Portraits are hand-built editorial vector illustrations keyed to each figure’s recognisable features — real photographs cannot be embedded in a generated file.
How to read this.(Base-case reading — the header percentages and cards above re-scale live with the simulator.) Modi's number is the joint probability that the NDA wins (~94% of simulations) and he takes a fourth term at 78. Every point that age or the RSS's informal 75-year norm removes from his row lands almost entirely on Shah, Adityanath, Gadkari, Rajnath, Chouhan and Fadnavis — not on the opposition. With the BJP alone at even odds of crossing 272, the live uncertainty in Indian politics is succession inside the BJP, not the contest between blocs. Rahul Gandhi's ceiling is the ~4% INDIA path, of which he holds most but not all; the regional names survive only in the ~7% of worlds without an NDA majority. Rows showing <1 are alive but below half a percent.