Prediction Models

Our Prediction Models

Learn about the different models we use to predict NCAA football games

Ensemble Model

Most AccurateCreated 2021

Our Ensemble Model combines predictions from multiple machine learning models, including ELO ratings, advanced statistics, and historical performance data. By weighing the outputs of each model based on their historical accuracy, the Ensemble provides our most reliable predictions.

72%

Season Accuracy

76%

Last Week

+8.2

Units (ATS)

65%

O/U Accuracy

Key Features:

Combines multiple predictive algorithms
Automatically adjusts model weights based on recent performance
Factors in team-specific metrics that other models miss
Best model for spread betting and game outcome predictions

Advanced Stats Model

Best for TotalsCreated 2022

Our Advanced Stats Model focuses on efficiency metrics rather than raw statistics. It considers factors like yards per play, success rate, havoc rate, and explosive play percentage to create a more nuanced prediction of game flow and scoring.

68%

Season Accuracy

71%

Last Week

+6.5

Units (ATS)

72%

O/U Accuracy

Key Features:

Focuses on efficiency and pace metrics
Strong predictor of game totals (over/under)
Accounts for opponent strength and schedule adjustments
Best model for identifying mispricings in total markets

Power Rating Model

Best for UpsetsCreated 2020

The Power Rating Model assigns a numerical value to each team based on their performance, creating a normalized scale that allows for direct comparison. This model is particularly strong at identifying undervalued teams and potential upsets.

65%

Season Accuracy

68%

Last Week

+4.8

Units (ATS)

62%

O/U Accuracy

Key Features:

Normalizes team performance on a single scale
Excellent at identifying line value and mispriced underdogs
Factors in home field advantage by venue
Best model for finding value on underdog moneylines