EVENT FORECASTS
Every big event, broken down to the last segment.
Elections and major events are researched in depth, then split into segments, regions and demographics. Reviewed by a person, refreshed every day.
THREE LAYERS OF DETAIL
Start with the headline, drill into the rest
Segments
Every option or candidate tracked, ranked by projected share or chance from 0 to 100. Up to 60 can be listed.
Regions
The same projection by geography, so you can see where a result is being won or lost.
Demographics
How support splits by group for the top two or three segments. Only included when genuine polling data exists. Never guessed.
LIFE OF A FORECAST
Three states, always visible
PENDINGBeing researched
A forecast request has been created and the AI is searching for real data. It stays here until research returns enough to fill it in.
PUBLISHEDLive and refreshed
Reviewed and approved by a person. From here it is refreshed in place every 24 hours and never drops off the site during an update.
INSUFFICIENT DATANot enough to say
After 5 research attempts with no real data, it stops retrying and is flagged for a person to look at, rather than showing made-up numbers.
OUR RULE
If we can't find real data, we say so.
Forecasts are filled in only from what the research actually finds, grounded in live web search. When something can't be sourced, such as a demographic split nobody has polled, that part is left out instead of estimated. A smaller honest forecast beats a bigger invented one.
HOW THE AI ANALYST COLLECTS DATA
How a forecast is researched
Each forecast is researched from live sources, not generated from memory.
WHERE THE DATA COMES FROM
Live web search
Google Search grounding finds current polling, reporting and official figures for the event.
Published polls and surveys
Used for segment, regional and demographic splits. Only genuine data is used.
Administrator review
A person checks each forecast before it is published.
WHAT THE ANALYST DOES, STEP BY STEP
- 1
A forecast is requested
An administrator creates a request, which starts as pending.
- 2
The AI researches it daily
It searches for real data and returns a headline stat, a summary and ranked segments, regions and demographics.
- 3
Retries if data is thin
If research finds nothing real, it tries again on later cycles, up to 5 attempts.
- 4
Flags what can't be sourced
After 5 failed attempts, it is marked insufficient data and handed to a person instead of being filled with guesses.
- 5
Refreshes once live
Published forecasts are updated in place every 24 hours.
Sourced or left out
Any section the research can't support, such as demographic splits without polling, is left out instead of estimated.
Limitations
- Forecasts summarise polling and reporting available at the time of the last refresh. Polls can be wrong.
- Segment, region and demographic detail depends on what data exists and varies from event to event.
- Forecasts are context for event risk. They are not trading signals or recommendations. See the Risk Disclosure.
QUESTIONS
Event Forecasts, answered
A researched tracker for a specific election or event. It has a headline figure, a short summary, and a ranked breakdown by segment, region and (where data exists) demographic group.
The AI researches each one using live web search, then an administrator reviews it before it is published. Published forecasts are then refreshed automatically.
Once published, a forecast is refreshed in place every 24 hours, so it never disappears from the site while it updates.
Segments are the options or candidates being tracked and their projected share. Regions break the projection down geographically. Demographics show how support splits by group, and are only included when genuine polling or survey data exists.
If research keeps coming back without real data (a purely hypothetical scenario with no polling, for instance), the forecast is marked 'insufficient data' after 5 attempts instead of retrying forever, and a person takes a look.
No. They summarise available polling and reporting at a point in time. Real outcomes can differ.
They are context for event risk, shown beside the rest of SaPEX NEXUS. They are not trading signals.
It runs live web searches for the event, looking for published polls, reporting and official results, then fills in a headline figure, a summary and ranked segments, regions and demographic splits.
The forecast reflects what the research found at that time, and a person reviews it before it is published. Daily refreshes can change the figures as new data appears.
Only from genuine polling or survey data found in research. If none exists, that section is left out entirely.
Forecasts are created by SaPEX NEXUS administrators. Each starts as a pending request that the AI researches.