Training camps are opening, rosters are close to set, and the model has enough current data to start forming a real picture of the 2026 NFL season
How does the AI model predict NFL teams before the 2026 season starts?
Before real 2026 games are played, the model starts from each team's end-of-2025 rating, regressed toward league average to account for normal year-over-year variance, then adjusts for offseason changes — free agency and draft additions weighted heavily by position value, with quarterback changes carrying the most weight, plus a separate adjustment for coaching and scheme changes. These early-season ratings carry wider uncertainty than in-season predictions and get sharpened as training camp, preseason injury news, and eventually real regular-season games provide current data.
With training camps opening across the league, the 2026 NFL season is close enough that our model has meaningful current-season data to work with — offseason roster moves, coaching changes, and draft additions are largely finalized, even if preseason performance and final roster cuts will still move the numbers over the coming weeks.
This is an early preview rather than a settled forecast. NFL team ratings move a lot between now and Week 1 as preseason injuries, camp battles, and final roster decisions play out, and the model will keep updating through August as that information comes in.
How the Model Builds Early-Season Ratings
Before real 2026 game data exists, the model leans heavily on a blend of the team's rating carried forward from the end of the 2025 season (regressed toward league average to account for typical year-over-year variance), adjusted for offseason personnel changes weighted by position value — a starting quarterback change moves the rating far more than a depth-chart change at most other positions.
Coaching changes get a separate adjustment layer, since a new offensive or defensive coordinator can shift a team's scheme and effectiveness independent of the underlying roster talent, though the model treats this adjustment with wider uncertainty than a roster-based change, since new-scheme performance is inherently harder to project before real snaps are played.
What the Model Is Watching Through Camp and Preseason
A few specific signals meaningfully move NFL ratings between now and the season opener: quarterback competition outcomes where a starting job is genuinely contested, significant camp injuries (particularly to offensive line and quarterback, the positions where an injury most directly affects team-level rating), and roster cut decisions that clarify a team's actual depth chart versus its offseason additions on paper.
Preseason game performance itself gets relatively little direct weight in the model, since starters typically play limited snaps and the games are widely understood across the league as evaluation tools rather than competitive results — but the injury and depth-chart information that comes out of camp and preseason carries real weight.
Reading Early-Season AI Predictions Responsibly
Early-season NFL predictions carry meaningfully more uncertainty than in-season predictions do, simply because there's far less current data to work from — a team's actual 2026 performance hasn't happened yet, unlike a prediction made in Week 10 with nine games of current-season data available. Treat preseason and early-season ratings as a reasonable starting estimate that will sharpen considerably once real games are played.
This is also the point in the calendar where a team's rating is most likely to diverge from public perception, since public narratives often lag behind actual roster and coaching changes — which is exactly the situation where checking the model's numbers against current market pricing (win totals, division odds) is most useful for spotting where perception hasn't caught up to the underlying data yet.
Conclusion
The 2026 NFL season model is built on last season's baseline, adjusted for the roster and coaching moves that have happened since, with wide uncertainty bands that will narrow considerably as camp battles resolve and real games are played. Check back through August as preseason data comes in, and follow our full model output on the American football predictions page once the regular season begins.
Frequently Asked Questions
How accurate are NFL predictions made before the season starts?
Less accurate than in-season predictions, since there's no current-season game data to work from yet — the model relies on last season's baseline plus offseason changes. Accuracy typically improves considerably a few weeks into the regular season once real current-form data is available.
What NFL roster changes matter most to an AI model's ratings?
Starting quarterback changes carry the most weight of any single position change, given the position's outsized effect on team performance. After that, offensive line and pass rush changes tend to have the largest effect on team-level ratings among non-quarterback positions.
Does preseason performance affect AI predictions for the regular season?
Only indirectly. Preseason results themselves get little weight since starters play limited snaps and results are treated leaguewide as evaluation exercises, but injury news and roster-cut decisions that emerge during camp and preseason do meaningfully affect a team's rating heading into Week 1.
