One underdog, one rule, tracked all season. This week the model's math points to Atlanta at Green Bay — and grades it thin.
NFL Dogs picks exactly one moneyline underdog every week of the season. Not three, not a parlay, not a card of leans — one dog. The selection isn't a gut call. Every week, the model runs its numbers on every game, and among every underdog it has a play on, this column takes the single one with the largest edge over the market's break-even price. That's the whole rule. It doesn't change week to week, and it doesn't bend for a team you like better or a storyline you find more interesting.
The record is tracked in the open, win or lose, and right now it's not pretty. That's part of the point of doing this in public.
The play is Atlanta Falcons moneyline at +200, on the road at Green Bay. A price of +200 needs to win only 33.3% of the time to break even. The model puts Atlanta's actual win probability at 37.5%. Subtract one from the other and you get a 4.2 percentage point gap — call it a 4.19% edge — and that gap is the entire justification for the pick. There's no second layer to it.
For context, the market has Atlanta as a 4.5-point underdog with the game total set at 42.5. Those are just the market's numbers, laid out so you can see what the model is arguing against. The model isn't saying Atlanta wins this game. It's saying that a team given a 37.5% chance to win is being priced by the market as if it were worse than that — a 33.3% shot or lower. A 37.5% team still loses more often than it wins. That's the nature of an underdog. The claim here is about price, not outcome.
This week's play carries a D grade from the model. That's a low-confidence rating, near the bottom of what the model will actually put a play on. It is not a typo and not false modesty — it means the model itself sees this as one of its weaker plays of the year, not a favorite, not a system best bet, and definitely not a lock. An underdog column is obligated to take the best dog on the board every single week, and some weeks the best dog available is still a thin one. This is that week. Nothing about a +200 price or a 4.19% edge upgrades a D into something it isn't.
Two weeks in, this column is 0-2. Week 1 took Carolina at +140 and lost. Week 2 took Arizona at +180 and lost. That's the entire sample — two plays, two losses, nothing more to read into it yet in either direction.
This is one model's output being tracked over a small number of games, not a proven system and not a betting recommendation. A 37.5% win probability means Atlanta is expected to lose this game more often than it wins it. The only argument being made is that +200 undersells that probability by more than it should. Whether that argument holds up gets decided on the field, and the column will report it exactly as it happens, same as it has the first two weeks.