The home edge is shrinking
From 2000 to 2009 home teams won 57.4% of NFL games played at a home stadium. Since 2016 the share is 55.1%. The low point was 2020 at 50.4%, the season most games were played without crowds.
Season by season
The best season for the model was 2009 (Brier 0.205, 69.3% of winners); the worst was 2006 (Brier 0.235, 58.4%). A Brier score of 0.25 is what always saying 50 percent would earn, so every season sits below it, but not by much: football games are close to coin flips more often than records suggest.
| Season | Games | Winners picked | Brier | Home win share |
|---|---|---|---|---|
| 2025 | 285 | 63.9% | 0.224 | 53.6% |
| 2024 | 285 | 69.1% | 0.213 | 53.6% |
| 2023 | 285 | 60.4% | 0.233 | 56.6% |
| 2022 | 284 | 63.4% | 0.222 | 57.4% |
| 2021 | 285 | 58.3% | 0.234 | 51.4% |
| 2020 | 269 | 65.4% | 0.217 | 50.4% |
| 2019 | 267 | 64.0% | 0.224 | 52.3% |
| 2018 | 267 | 62.9% | 0.221 | 60.1% |
| 2017 | 267 | 65.2% | 0.217 | 57.5% |
| 2016 | 267 | 64.0% | 0.217 | 58.6% |
| 2015 | 267 | 65.9% | 0.225 | 54.0% |
| 2014 | 267 | 68.9% | 0.206 | 58.2% |
| 2013 | 267 | 65.9% | 0.214 | 60.3% |
| 2012 | 267 | 62.2% | 0.219 | 57.6% |
| 2011 | 267 | 63.7% | 0.213 | 57.7% |
| 2010 | 267 | 62.2% | 0.233 | 55.3% |
| 2009 | 267 | 69.3% | 0.205 | 57.7% |
| 2008 | 267 | 64.0% | 0.217 | 57.0% |
| 2007 | 267 | 64.8% | 0.215 | 57.4% |
| 2006 | 267 | 58.4% | 0.235 | 54.1% |
| 2005 | 267 | 67.8% | 0.211 | 58.1% |
| 2004 | 267 | 64.0% | 0.223 | 56.8% |
| 2003 | 267 | 64.4% | 0.220 | 61.5% |
| 2002 | 267 | 59.9% | 0.226 | 58.8% |
| 2001 | 259 | 65.6% | 0.229 | 55.4% |
| 2000 | 259 | 63.3% | 0.222 | 56.6% |
How the backtest works
Ratings start at 1505 for every team in 1999, so that season only serves to warm the system up and is left out of the averages. From then on each game is forecast from the ratings as they stood before kickoff, including a home-field bonus of 65 Elo points, and the ratings update afterward. Nothing is refitted with hindsight. The constants and formulas are on the methodology page, and the Elo guide explains why they were chosen.
Frequently asked questions
How often does the Elo favorite win NFL games?
Across 7,017 games from 2000 to 2025, the team with the higher pregame Elo probability won 64.1% of the time. The rate moves between seasons; the table on this page lists every one.
What is a good Brier score for NFL game predictions?
Always saying 50 percent scores 0.25. The site’s Elo averaged 0.221 over these seasons. Scores near 0.20 are good for pregame NFL forecasts, because many games are close to coin flips.
Has home-field advantage in the NFL declined?
Yes. Home teams won 57.4% of games from 2000 to 2009 and 55.1% from 2016 on. The lowest season was 2020, when most games were played without crowds, at 50.4%.
Can I use the home-field chart on my own site?
Yes. The chart is published as a standalone SVG with an embed snippet on this page. The underlying game data is nflverse, licensed CC-BY 4.0, and the snippet carries that credit.