How to read an NFL prediction

An EdgeIQ game page shows several numbers at once. They are not competing opinions — each one answers a different question about the same matchup.

Win probability: how often, not whether

Win probability is the share of similar matchups in which the favored team comes out ahead. Read 68% as 'about two times in three', never as 'this team wins'. Probabilities below roughly 60% describe a close game, not a weak team.

Projected score: the central estimate

The projected score is the middle of a range of plausible results, not a forecast of the exact final. NFL scores routinely land a touchdown or more away from any model's central estimate. Model Lab publishes the typical size of that miss so you can judge how tight the range really is.

Projected margin: the most useful single number

Margin is the projected difference between the two teams. It is the number that lines up most directly with market spreads, which makes it the easiest way to spot where the model and the market disagree.

Confidence: how much weight to give it

Confidence reflects how strongly the underlying data separates the two teams. A high-confidence call on a matchup with thin data is worth less than a moderate-confidence call on two teams with a full season of results behind them, which is why EdgeIQ shows sample size next to every confidence figure.

Three habits that make the numbers useful

• Read the projection with recent form, injuries and rest, never in isolation
• Treat a run of misses inside a confidence band as normal unless the sample is large
• Write down what you expected before kickoff so you can grade yourself, not just the model