Behind the Data.
An independent election forecasting project built around one simple idea: using early voting data to understand where an election may be heading before Election Day.
Hi, I'm Harry.
I started EDG in 2024 with the intention of developing a predictive analysis tool capable of projecting the outcome of major U.S. elections using the information available through the early voting system.
What began as a personal project has gradually developed into something much bigger — combining data, historical voting patterns and new information as it becomes available.
Built independently
EDG is independently developed and maintained. The model, website, data systems and analysis are built and managed by me, with the project continuing to evolve with every election cycle.
Built around data
Rather than relying solely on polling, EDG focuses heavily on actual voting behaviour — particularly early and mail voting — and compares new information against historical election data.
What is EDG?
Election Data Guy is an independent election forecasting project tracking early voting, mail voting and turnout to analyse the direction of elections across the United States.
📊 Data first
EDG takes historical early vote information and compares it with the latest figures coming from states and counties. The objective is to identify patterns that can provide an indication of how voting behaviour is developing.
🇺🇸 State by state
Elections are not treated as one national dataset. EDG analyses states individually, taking into account their own voting systems, historical behaviour and available early voting information.
🔄 Constantly evolving
Every new election provides another opportunity to test the model. New results are compared against previous cycles, allowing the methodology to be refined over time.
🔎 Transparency
EDG publishes the underlying indicators used to explain its projections wherever the relevant data is available, allowing users to see how a projection is developing rather than simply seeing a final call.
How the model works.
The EDG model combines historical voting behaviour with new early voting data to produce an indication of where each race may be heading.
Historical data
Previous election results establish the historical baseline for each state and, where possible, county.
New voting data
Fresh early and mail voting figures are added as states release new information.
Comparison
Current voting patterns are compared with historical patterns and the characteristics of previous elections.
Projection
The resulting indicators are used to produce a projected range and identify the direction of the race.
Testing the model.
EDG is designed to be tested against real election results. Every election provides another dataset through which the model can be evaluated and refined.
Not every state tells us the same story.
One of the biggest challenges of analysing early voting nationally is that states collect and publish different information.
Different systems
States have different rules, reporting systems and definitions surrounding early and mail voting. This means national comparisons cannot always be made on a perfectly like-for-like basis.
County-level analysis
Some states do not publish party registration or partisanship alongside voter files. Where this happens, EDG can use county-level information and historical election results to provide additional context.
Unavailable data
Some states restrict the release of early voting information, while others may not provide enough information for EDG to produce a meaningful projection.
Every projection has limits
Early voting is an indicator — not a prediction of an individual voter's eventual choice. EDG therefore treats its outputs as model-based projections rather than guarantees of an election result.
“The goal isn't simply to predict an election. It's to understand how the election is developing.”
— Election Data GuyWhy EDG exists.
What started as a passion project has evolved into an independent
election data platform.
The purpose of EDG is to make complicated election data easier to
understand. Instead of waiting until Election Day to see the final
result, the project follows the information that becomes available
during the weeks beforehand.
The model will never be finished.
Every election creates new data, new circumstances and new ways
to test the methodology. That is what makes EDG an ongoing project
rather than a one-time forecast.
The ambition is simple:
build a model that gets better every time it is tested.
Model projections
EDG projections are generated using the project's methodology and available data. They should not be interpreted as guarantees of election outcomes.
Data availability
The availability, quality and timing of early voting data varies between states. A state may therefore be excluded when the available information is insufficient for the model.
Follow the data.
Explore the latest early voting figures, state projections and election analysis from Election Data Guy.
Explore EDG →