ESPN → LOCAL DATABASE
Schedules, rankings, records, venues, kickoff times, networks, and final scores come from ESPN. Pages read the local COFO database, so a temporary ESPN outage does not take the site down.
OPENING THE BLACK BOX
HOW COFO
THINKS.
Explainable football math. No secret injury reports. No vibes disguised as science. No pretending a 96% pick is a lock.
MODELSATURDAY SCIENCE v1.0
GRADED PICKS0
WINNER ACCURACYWAITING ON FINALS
CORRECT0
THE SHORT VERSIONWHO HAVE YOU PLAYED,
WHO HAVE YOU PLAYED,
AND WHAT DID YOU DO TO THEM?
COFO builds a power rating from completed games, adjusts it for opponent quality, blends each offense with the defense across from it, and turns the projected margin into a win probability. Then Saturday gets to grade the homework.
CONTROL THE BLOWOUTS
Winning by 55 matters, but it should not break the calculator. Scoring margins are capped at 35 points before ratings are built.
REGRESS THE SMALL SAMPLES
One wild Saturday is not a personality. Early scoring averages are pulled toward the FBS scoring mean using two games of prior weight.
ADJUST FOR WHO YOU PLAYED
COFO runs twelve opponent-adjustment passes. Beating a strong team helps more than flattening a cupcake; surviving a brutal schedule earns credit.
RESPECT RECENT FORM—A LITTLE
The last three games influence 25% of the margin profile. Hot streaks matter, but the model does not forget the rest of the season overnight.
PROJECT THE MATCHUP
Each team's scoring offense is blended with the opponent's scoring defense. Power rating, schedule strength, recent form, and a two-point home-field edge adjust the projected score. Neutral-site games get no home bonus.
MARGIN → WIN PROBABILITY
A logistic curve turns projected margin into probability, bounded between 5% and 95%. College football always reserves the right to become ridiculous.
FREEZE THE RECEIPT
Published predictions are saved snapshots. When the game goes final, COFO grades the original pick and margin—it does not rewrite history.
LAST YEAR HELPS.
THIS YEAR TAKES OVER.
Final opponent-adjusted ratings and scoring profiles from the previous season begin at 65% weight. That carryover fades after every new game and reaches zero after three. AP rankings add a modest extra clue, never a commandment.
LOW IS AN HONEST ANSWER.
No data, fewer than three games, or a projected margin under three points produces LOW confidence. Five games plus a seven-point edge can earn HIGH. Everything between is MEDIUM.
THE FUN TRANSLATION LAYER
WHAT THE BADGES MEAN
An unranked pick over a top-15 team, a ten-plus ranking gap, or a serious power-rating rebellion.
COFO backs the team that rankings or its own power rating consider the underdog.
The predicted winner is lower-ranked—or unranked against a ranked opponent.
At least 82% win probability or a projected margin of 17-plus.
At least 68% win probability or an eight-point projected edge.
A projected margin of 2.5 points or less. Deep breaths.
Projected margin between 2.5 and seven points.
The permanent postgame receipt. No appeals.
COFO is an entertainment model, not betting advice. It does not know injuries, suspensions, weather changes, depth-chart news, or whether a mascot has achieved dangerous levels of motivation.
SEE THE PICKS →