Some 2026 predictions may look strange due to limited data.
Model Version: 2025
The model utilizes inputs of both historical data for each district or state, as well as recent polling. Certain bonuses and penalties are applied for a variety of factors to produce a final result.
This is an upgrade from the 2024 model, which failed to accurately predict the winner of the 2024 presidential election. This upgrade has also enabled a district-by-district analysis for the US House for the first time.
With this upgraded model, confidence intervals have been introduced. The model was run over several recent elections to build that confidence range. These intervals are different for each race type and help determine when a race moves from "Toss-Up" to "Lean", "Likely" or "Solid".
Tests for the new 2025 model did not include "confidence" ranges - a major factor used in determine accuracy. These tests were used to inform the confidence ranges for the 2025 model. As such, there is limited information on exactly how accurate this new model will be. The 2025 and 2026 elections will be major tests for the 2025 version. Because the only major (regular) races in 2025 are for the governors of New Jersey and Virginia, the only change to the 2026 model will be a minor adjustment to the confidence ranges for gubernatorial races based on the 2025 results.
However, removing the confidence ranges from the conversation, the model performed more accurately than the 2024 version.
The model is informed largely by historical results and recent polling data. As such, early in a cycle (especially pre-primary), limited polling means that the model relies heavily on historical data. This is why some races, like Montana's US Senate elections, may show up as something closer than you'd expect - there are still recent examples of Montana voting for a Democrat for US Senate (see 2018), which still have an echo of an impact on the model.Â
As polls start to come in, you will likely see those races shift to something you are more likely to expect.
Honestly? I'm just some guy from Massachusetts.
Originally from Iowa, elections have always interested me. Now I'm trying my hand at predicting them.
Now get ready to go vote!