Methodology
3NG combines college football data with independently developed ratings, rankings, and prediction models. This page explains how those systems fit together and what the numbers shown throughout the site represent.
Foundation
3NG uses outside data sources as inputs to the system. Those inputs include information such as teams, conferences, schedules, game results, statistics, recruiting and roster information, betting markets, and other game-related data.
A significant portion of the college football data used by 3NG is provided through CollegeFootballData.com. The source data is collected and organized within the 3NG database system before being used by the site's ratings, rankings, and prediction models.
Outside data is an input to 3NG. The ratings, rankings, model calculations, and predictions presented as 3NG results are produced independently by the 3NG system.
College football data provided in part by CollegeFootballData.com.
Team Evaluation
Ratings are the numerical building blocks of the 3NG system. Rather than relying on a team's record alone, 3NG evaluates different parts of team performance and then combines those measurements as the season develops.
Offensive Rating measures overall offensive performance using multiple efficiency components, including success rate, predicted points added, explosiveness, scoring opportunities, short-yardage performance, field position, and passing and rushing efficiency.
The components are combined and standardized within each season. A rating of 100 represents the season average, with higher values representing stronger offensive performance.
Defensive Rating evaluates how effectively a defense limits the same types of efficiency and production measured on offense, including success, explosiveness, scoring opportunities, field position, and passing and rushing efficiency.
Defensive performance is standardized within each season and placed on the same 100-centered rating scale used by the offensive system.
Special Teams Rating measures kickoff returns, punt returns, kickoff coverage, punt coverage, and return-touchdown production and prevention.
Those components are standardized and combined to measure the field-position and scoring value created or surrendered by special teams.
Strength of Schedule measures the quality of the opponents a team faces rather than simply using opponent win-loss records.
Opponent quality is built from 3NG offensive, defensive, and special-teams ratings and also considers the strength of the opponents' schedules. The system produces separate ratings for games already played, games remaining, and the complete schedule.
Before current-season performance is available, preseason opponent quality incorporates previous performance along with returning production, recruiting, roster talent, and transfer information.
Strength of Record asks a different question from Strength of Schedule: how difficult would it have been for an elite team to produce the same record against that schedule?
3NG uses a Top-5 caliber benchmark team and estimates its win probability for each completed game based on opponent quality and game location. Those individual probabilities are combined to estimate the likelihood of the benchmark team matching or exceeding the team's actual number of wins.
Better records against difficult schedules therefore receive more credit than identical records produced against weaker competition.
Team Efficiency combines offensive and defensive efficiency with drive performance, third-down and fourth-down efficiency, scoring-opportunity efficiency, and special teams.
The result is a broad measure of how efficiently a team performs across the situations that determine possessions and scoring.
3NG maintains an Elo rating that updates as games are played. Expected results are based on the rating difference between opponents with an adjustment for home-field advantage.
Ratings move after each result based on the expected outcome and actual result, with margin of victory included in the update. Ratings carry from one season to the next with offseason regression rather than completely resetting every year.
Team Strength is the broadest 3NG team rating. It combines Team Efficiency, Offensive Rating, Defensive Rating, Strength of Record, Strength of Schedule, Elo, conference strength, and Special Teams Rating.
Because it draws from both performance and schedule context, Team Strength is designed to represent the overall quality of a team rather than one individual area of performance.
National Comparison
A rating is a numerical evaluation. A ranking is the ordering produced when those evaluations are compared across teams or conferences.
The 3NG Team Power Rankings combine three broad areas: current performance, résumé, and underlying team strength.
Current performance emphasizes Team Efficiency while also considering offense, defense, and special teams. Résumé incorporates Strength of Record, Strength of Schedule, record quality, and results against different levels of competition. Underlying strength combines Team Strength and Elo.
Game location also matters when evaluating résumé results, with road wins receiving more credit and home losses carrying a greater penalty.
The Simple Rating System evaluates teams using scoring margin and recursively adjusted opponent strength.
The system solves team and opponent strength together so that performance against strong competition carries different meaning than the same scoring margin against weak competition. SRS is centered at 0, representing an average FBS team, with the rating expressed in estimated points above or below average.
Conference Power Rankings measure both the internal quality of a conference and how that conference performs outside its own membership.
Internal quality considers the average, median, upper-quarter, and lower-quarter strength of the conference. Non-conference performance is adjusted for opponent quality and game location.
Before in-season Team Power Rankings are available, 3NG uses a separate preseason model to estimate expected team strength entering Week 1.
The preseason model begins with established program strength and incorporates recent Team Strength, recruiting and roster talent, previous-season defensive and special-teams performance, returning production, replacement quality, transfer-portal additions, quarterback experience and performance, coaching, roster continuity, and championship-level carryover.
Offseason adjustments are intentionally limited so that new information can move a team's preseason expectation without completely overriding multiple seasons of established performance.
Game Forecasting
3NG game predictions use team ratings, game information, historical performance, betting-market information, and other available game context to estimate the expected outcome of a matchup.
The production system does not rely on one model for every prediction target. Different models are used for winner probability, expected margin, and expected total based on their historical performance.
The official win probability is produced by the 3NG Meta Gate model. The model learns when different prediction sources have historically performed best and dynamically blends those sources for each game.
The team with the higher resulting win probability is listed as the predicted winner.
Official projected margin and projected total come from the 3NG market-residual model. Rather than ignoring the betting market, the model evaluates where 3NG's football information indicates the market expectation may be high or low.
The projected margin and total are then combined to derive the projected score for each team.
Confidence reflects the strength of the model's predicted game outcome. It is intended as a quick indication of how strongly the prediction system favors its selected winner, not as a guarantee of the result.
For current production predictions, 3NG compares its projected margin and total with available market lines to identify differences between the model and the betting market.
These views are model comparisons against the market and should not be interpreted as guaranteed betting outcomes.
Historical
Predictions shown for the 2017 through 2025 seasons are historical model outputs used to evaluate how the 3NG prediction system performs against completed games.
These predictions were not publicly issued by 3NG before those games were played and should not be represented as live historical picks.
Production
Beginning with the 2026 season, official 3NG predictions are generated before games are played and stored as timestamped production prediction snapshots.
These are the live predictions used by the website and subsequently graded when final game results become available.
Model Evaluation
Prediction performance is measured against completed FBS-vs-FBS games.
3NG does not present historical ATS or Over / Under performance for 2017–2025 because a verified historical source suitable for representing those results as 3NG betting performance is not currently available.
Models and rating systems may continue to evolve as the 3NG system is developed. Methodology descriptions on this page are intended to reflect the production systems used by the site.