NFL 2026 · Week 3
Kick-off
Fri 25 Sep, 10:15am AEST
Thu 24 Sep, 8:15pm ET
AU consensus as of 24/09/2026, 4:05:59 pm
Power Comparison
ATLRating -5.5
Defence by zone (EPA allowed)
GBRating -1.0
Defence by zone (EPA allowed)
Zone cells are EPA/play vs league mean, left to right in model order. Green = better for that team. Hover a cell for the zone name.
Matchup Edges
ATL offence vs GB
deep-left passing
+0.33 EPA/play
GB offence vs ATL
short-right passing
+0.13 EPA/play
GB offence vs ATL
deep-middle passing
+0.25 EPA/play
Market
Spread (GB)
Open -6.5 → Now -5·Move +1.5
AU bestGB -4.5 @ $1.87 SportsBet
AU best (other side)ATL +5.5 @ $1.95 Dabble AU
Global bestGB -4.5 @ $1.98 LowVig.ag
Total
Open 44.5 → Now 43.0·Move -1.5Moved toward model
AU best overOver 42.5 @ $1.85 TAB
AU best underUnder 43.5 @ $1.90 PointsBet (AU)
Global best overOver 24.5 @ $1.03 Courtside
AU best ATL ML
$3.10 Unibet
11 AU books · 33 total · 9 snapshots
The Call
QB situation unresolved - no rating basis
No rating basis
ATL: 2026-09-24: Michael Penix Jr. is confirmed to start at Green Bay, but the Penix return adjustment is not approved. Keep Atlanta-Green Bay unrated until Shaun signs off on a numeric QB value.
Spread · Rating withheld
Unrated
QB UNRESOLVED / UNRATEDTotal · Rating withheld
Unrated
QB UNRESOLVED / UNRATEDQuarterback Michael Penix Jr. has been confirmed to start at Green Bay for Atlanta, but the internal adjustment for his return hasn't been signed off, leaving the quarterback situation as the whole story in this one. The model cannot price Atlanta's quarterback situation and has no rating basis for this game; its margin of 7.1 points and total of 37.1 are shown only as context, not a read on the market.
Even setting the quarterback issue aside, this early in the season the rating for both teams is still last year's numbers blended with the market's post-offseason view of each side (the market anchor) and scaled by how much of each roster carried over (continuity), which is exactly why a QB change like this one can't simply be folded in without sign-off. The starter announcement and any approved adjustment for Penix are what turns this from an unrated game into a priced one.
Why · QB situation unresolved - no rating basis (ATL)
Situational
2025 season · FTN charting via nflverse · min 40 snaps per cell (25 for box counts)
ATL · 0 dropbacks
Starter not named
QB situation unresolved: no starter to profile
GB · 461 dropbacks
Jordan Love
Clean pocket EPA/db+0.54 (332 snaps)
Under pressure EPA/db-0.54 (129 snaps)
Pressure rate faced28%
Int-worthy rate2%
Drop rate (catchable)5%
QB-fault sack rate1%
League: +0.26 clean · -0.46 pressured
ATL defence: pressure 31% · blitz 35%
ATL offence vs GB coverage
| Shell | GB usage | ATL EPA/db | League |
|---|
| Cover 3 | 34% (204 snaps) | -0.06 (154 snaps) | +0.06 |
| Cover 2 | 28% (171 snaps) | +0.17 (140 snaps) | 0.00 |
| Cover 1 | 17% (101 snaps) | -0.13 (136 snaps) | +0.06 |
| Cover 4 | 10% (59 snaps) | +0.17 (56 snaps) | +0.07 |
GB man coverage 20% · ATL vs man -0.11, vs zone +0.06
GB offence vs ATL coverage
| Shell | ATL usage | GB EPA/db | League |
|---|
| Cover 1 | 28% (176 snaps) | +0.50 (132 snaps) | +0.06 |
| Cover 3 | 25% (159 snaps) | +0.25 (141 snaps) | +0.06 |
| Cover 2 | 16% (99 snaps) | +0.06 (130 snaps) | 0.00 |
| Cover 4 | 16% (99 snaps) | +0.36 (55 snaps) | +0.07 |
ATL man coverage 33% · GB vs man +0.47, vs zone +0.16
ATL tendencies
Play action rate20% · +0.23 vs -0.05 EPA (117 snaps)
RPO rate5% (50 snaps)
Screen rate9% (50 snaps)
Personnel11 46%, 12 38%, 21 9%, 13 7%
Def pressure / blitz / man31% / 35% / 33%
Box vs run, EPA allowedlight +0.03 (228 snaps) · neutral -0.03 (155 snaps) · heavy -0.25 (34 snaps)
GB tendencies
Play action rate26% · +0.40 vs +0.20 EPA (141 snaps)
RPO rate10% (94 snaps)
Screen rate7% (38 snaps)
Personnel11 61%, 12 33%
Def pressure / blitz / man31% / 21% / 20%
Box vs run, EPA allowedlight -0.07 (222 snaps) · neutral -0.07 (177 snaps) · heavy +0.17 (45 snaps)
Model Form
2025 season, weeks 6-17, V3 at best available line: 29-18, +17.8% ROI
Weeks 1 to 5 are published analysis-first. Only 3,666 plays of 2026 play-by-play exist so far (weeks 1 to 2), so ratings are still mostly last season's numbers blended with the market's post-offseason view of each team and scaled by roster continuity, with the 2026 sample phased in as it grows. The model alone cannot see coaching, roster or scheme changes - only the market anchor, roster continuity and the manual override tables capture those. In last season's backtest the same approach went 32-46 against the spread and 41-37 on totals over weeks 1 to 5; 53% on totals is break-even at -110. Only edges of 10 points or more are staked (Play) before week 6; a 7-10 point edge is analysis only (Lean), not a tip.
Analysis only. Not financial advice. Gamble responsibly. Gambling Help Online 1800 858 858. Read our responsible gambling policy.
savvyplays.comPublished 24/09/2026V3.2