Model share at default settings · 20,000 primaries
Model resultloading default.json
Loading the precomputed default run.
Today's turning points
See one primary play outnot yet run
One primary plays out month by month from the current settings. The card shows who wins and why; expand it for the full timeline.
The field
Hover a card for the dossier · set In/Out to override the market
Poll average is an equal weight mean of national Democratic primary polls from the last 30 days. Harvard CAPS Harris samples all Democrats rather than likely primary voters and runs hot on Harris. Polls measure name recognition this far out. Markets drive the model. The gap is the story.
The turning point heatmap
What moves whom · 5,000 matched seed runs per levernot yet built
Click "Build heatmap" to compute how much each turning point swings each candidate. This shows what changes the odds, not just the odds themselves.
Turning point rankings
The election's most decisive variables · by average swingbuilt with the heatmap
Build the heatmap above and these rank automatically: the turning points that move the most outcomes, measured in average first-place swing across the field. This ranks the election itself, not the candidates.
The world
Basic shows the five biggest levers. Advanced reveals everything, including your bracket.
Candidate gates
66%
78%
60%
62%
55%
Silver Bulletin average Sep 10, 2026: Talarico plus 2.4. A win doubles his base strength and puts his run odds at 60.
52%
CNN Sep 2026: El-Sayed plus 3. Same shape as Texas.
90%
Environment
38%
Silver Bulletin average Sep 10, 2026: approve 38.5, net minus 19.5.
55%
CNN/SSRS Sep 2026: Jackson plus 3.
50%
35%
40%
35%
Share of a departing candidate's strength that moves to pool mates, primary pool at full weight and secondary at half.
50%
Your bracket
Set the run odds for the top six by market, then run the machine on your priorsmarket defaults loaded
The link encodes your six numbers. No accounts, nothing stored.
What moved
Since yesterday
Loading history.csv
THE 2028 MACHINE · A MODEL BUILT FROM ONE OPERATOR'S PRIORS, NOT A FORECAST · THE STRENGTHS ARE ARGUABLE ON PURPOSE; THE VALUE IS WHAT MOVES THE OUTCOME, NOT THE HEADLINE ODDS · PRIORS CURRENT TO SEPTEMBER 2026 · MARKET ANCHORS UPDATED NIGHTLY FROM KALSHI AND POLYMARKET · POLLS ENTERED BY HAND · POOL ASSIGNMENTS AND SHARES ARE OPERATOR PRIORS, NOT MEASURED VOTER DATA · THE DRISCOLL GLOBE · POLITICAL SYSTEMS DESK