How the model works

Our process is built to be repeatable and honest rather than clever. Here is what goes into every pick.

Two independent projections

We project each game two separate ways, using two well-regarded rating systems, and add a home-field adjustment. We only act when the two systems agree on which side has value. When they disagree, we pass. A single model running hot is how you talk yourself into noise.

Edge, and why it is not enough

The edge is the gap between our projected number and the market's number. But a points edge is worthless in a game destined to be a blowout, because the result lands nowhere near the spread. So we weight every edge by how competitive the game projects to be. We call the result the Chives Alpha Ratio: the edge divided by the game's blowout risk. A decisive edge in a projected one-score game is worth far more than a larger edge on a team getting three touchdowns. We fish competitive games and pass lopsided ones no matter how wrong the line looks.

The bar a pick has to clear

A pick must clear a minimum edge; anything smaller is inside the model's own margin of error. Most weeks only a handful of games qualify, and we do not pad the card to hit a number. A confidence score from 0 to 10 reflects the combined read, tempered for the uncertainty in each spot.

Closing line value: the real scorecard

Win or lose, the leading indicator of whether an approach is sound is whether it beats the closing number, the sharpest price the market produces. We grade every pick against the close, not just the final score, so the record measures process and not luck. It takes a full season for small edges to prove out.

What this is, and is not

This is analysis and commentary for informational and entertainment purposes. It is not betting advice, and it is not a guarantee. The edges are small and most bettors lose. Our numbers reflect the time of writing; odds and lines move. See the track record for how the picks are actually doing.