Make your Voter DNA
Mix and match traits like age and education to see how a person’s "DNA" determines if they lean Democrat or Republican.
Every percentage traces back to 2024 national exit polls - real vote breakdowns by age, race, education, religion, and more. Those margins were used to generate 60,000 synthetic voters via machine learning, each one built to reflect actual demographic voting patterns. A LASSO model was then trained on that dataset - it learned which traits meaningfully shift a vote, and quietly zeroed out the ones that don't. What survives is a precise, data-grounded weight for every trait option you see.
Each trait you select carries a learned signal - a pull toward Democrat or Republican and a measure of how strongly. Those signals are combined into a single probability. The National Baseline slider then adds a direct percentage-point shift on top, with no distortion.
Crucially, the model maps Interaction Deltas - meaning traits don't exist in a vacuum. Being "65+" might pull one way for a White voter, but differently for a Black voter. The Gravity Well splits these interaction synergies dynamically, so beams update in real-time based on your other selections.
Tap a beam - what the numbers mean?
Of every trait currently active, what slice of the total pull does this one contribute? Add a stronger trait and its share drops. Remove one and it rises. It's a live competition between your selections.
This trait's raw signal, as learned during ML training - how hard it pushes toward one party on its own, regardless of what else is selected. It controls the thickness and brightness of that trait's beam in the Gravity Well.
The central orb is your voter. Particles flow toward the blue pole (Democrat) or red pole (Republican) - denser flow means a stronger lean. Each beam is one active trait: thicker and brighter means higher pull strength, its direction shows which party it favors. Tap any beam to see that trait's stats. Tap the orb to see the full split.