Election Night Analysis: Art or Science
summary
Uses AEC polling-place XML and historical booth data fed into an ABC database (now running on Amazon) to predict seat outcomes by computing match two candidate preferred swings rather than raw totals, cutting variance to roughly one‑third to one‑half. Calls happen when a computed 99% confidence interval for a seat stays above 50% (with a 3% bottom-of-frame reporting threshold to avoid noise). Typical early overall margins show errors around ±3 seats; postals and rising pre‑poll voting are the main sources of delay and uncertainty.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:12Credits
- 3:39Australian Federal Elections
- 5:55Counting
- 7:18Indicating Preference Count
- 9:02Virtual Tally Room
- 11:22PostElection Counting
- 12:22LCA History
- 13:22The Statistical Problem
- 14:24Prediction
- 16:20Sources of Error
- 19:08Scatter Plots
- 27:01Election Night
related talks
01h 03mHow to Steal an Election
22mFacebook data and machine learning predicting elections?
44mAutomating Analysis
56mSports Analytics: Big Data in the Big Time
47mexpectation: A modern take on statistical A/B testing with e-values and martingales
21mLeveraging Multi-Armed Bandit Algorithms for Dynamic Decision Making
01h 17mLightning Talks
35mExposing Algorithms