WE NEVER PUNT.

WHAT THIS IS

NoPunt is a predictive analytics platform for NFL games. We publish tier-graded picks for every game on the schedule, every week, before kickoff. When the game ends, the result is logged whether we won or lost. The full ledger is public.

This is not a tout service. We don’t sell locks. We don’t promise guaranteed winners. We run a model, show you the output, and let the record speak.

THE MODEL

Three logistic regression classifiers trained on NFL play-by-play data going back to 2002. Each sub-model uses a different rolling window (24, 33, and 64 games). They vote. Majority wins. The ensemble processes 154 features including rolling EPA splits, home/road deltas, rest days, divisional indicators, QB adjustments, and the Vegas spread as a calibration anchor.

The output is a win probability, converted to a fair line and compared against the market. The bigger the gap between our number and the market’s number, the higher the confidence tier (S, A+, A, B, C). For a primer on reading those market numbers, see NFL betting odds explained. Full technical details are on the methodology page.

THE RECORD

Six seasons of tracked results. Every pick published before kickoff, every loss shown publicly, no retroactive edits. The disagree flag marks games where our model picked the opposite side of the consensus moneyline — those are tracked with real betting units. Browse the full results ledger or jump to this week’s picks.

THE PHILOSOPHY

Most sports prediction content operates the same way: post winners loudly, delete losers quietly, screenshot the hits, ignore the misses. NoPunt exists because that model is broken.

We publish everything. The 70% confidence pick that lost. The C-tier lean that somehow hit. The full card, not just the highlights. If a prediction model can’t survive public scrutiny of its complete record, it does not deserve your attention.

NOT A TIP SHEET

NoPunt is not a tipster, a handicapper collective, or a Discord full of “guaranteed plays.” It is a model with receipts. The differences:

• Every pick is generated by the same pipeline, not human hunches.
• Every result is logged automatically — we can't selectively delete losses.
• Confidence tiers tell you how strong the edge is before you decide anything.
• The methodology is public. You can audit how the picks are made.

THE NAME

“We Never Punt.” Always making a call, never hedging. Every game on the board gets a pick and a tier. No “stay away” labels, no dodging tough games. If the model has a lean, you see it.