Real-time forecasting
The real-time forecasting system feeds new events and data into the ontology to update its forecasts, records every forecast it makes, and corrects itself against actual outcomes.
Reflect each day’s events and keep rewriting the forecast
A forecast is not a report you publish once and forget. The field moves every day, and the forecast has to move with it.
New events are reflected right away. When news, opinion or indicators come in, the values of the affected factors in the ontology change, and Monte Carlo simulation rolls the impact thousands of times to produce a new distribution of outcomes. A Kalman filter separates the trend from the noise.
When there is something to ask, we ask the synthetic population right away. We pick only the people who match the conditions and correct the answers to the real population. No waiting days for a survey.
Every forecast is recorded. A forecast ledger keeps each forecast as it was made, and when results come in we compare them with reality and fix the next forecast. That is why it gets more accurate with use.
In Korea’s 9th nationwide local elections in 2026, this system came within 0.4 percentage points of the actual count in Seoul and called 14 of 16 metropolitan races.
See also
Last updated: