third angle · predict Delhi

waking the city…

PREDICTING

ask ↵

simulate and predict Delhi's electorate

The pitch

What if you could predict how a population responds to real events before they play out: an election, a ballot measure, a policy change?

We simulated Delhi's population, sampled across the demographics of all 7 Lok Sabha constituencies (Chandni Chowk, East Delhi, New Delhi, North East Delhi, North West Delhi, South Delhi, West Delhi), and asked it to predict a real election it had never seen.

Historically predictive

How do we know if the simulation is accurate? The honest test needs a model whose knowledge cutoff predates the event, so it can't have just memorized the outcome. For this leakage-free backtest we ran GPT-4o (knowledge cutoff October 2023) and asked the twin city to predict the May 2024 Lok Sabha election, a contest it had never seen. The live app you're using also runs GPT-4o.

Chandni Chowk · 2024 Lok Sabha

Of votes cast for the two leading candidates, what share went to the BJP (NDA) candidate rather than the Congress (INDIA bloc) candidate?

Actual: 54.7% NDAPredicted: 54.5% NDA

New Delhi · 2024 Lok Sabha

Of votes cast for the two leading candidates, what share went to the BJP (NDA) candidate rather than the AAP (INDIA bloc) candidate?

Actual: 54.7% NDAPredicted: 55.3% NDA

The future

All of decision-making is predicated on our understanding of causality. How will the choices we make influence the world? How will people react?

To enable Delhi, and then the country, to understand the causal nature of reality, we present third angle.

Built by Max Gabriel and Sourav Padhi.