NFL Week 3 2026 Predictions & Picks
Updated · market prices
Quick Answer
As of September 15, 2026, our Week 3 calls: 6 market favorites and 10 upsets straight up. Biggest upset call: Saints over Raiders at 67% to win (Kalshi 48¢). The model disagrees with the Kalshi price on 8 of 16 games; the widest gap is Steelers at 58% to win versus Kalshi's 48¢.
Next widest: Dolphins at 50% to win versus Kalshi's 42¢; Cowboys at 52% to win versus Kalshi's 45¢.
Every game below carries the model's win probability, the Kalshi price, and the gap between them — plus the straight-up winner and a projected score.
Calls by Benny Ricciardi and The 7 Oracles · DAEPA model
Thursday edge sheet: all 16 Week 3 games, our number vs Kalshi. Free.
No spam. Unsubscribe anytime.
- Falcons at Packers — Model winner: Packers (66%)
- Chargers at Bills — Model winner: Bills (65%) · The play: Bills ≤ 60¢
- Panthers at Browns — Model winner: Browns (55%) · The play: Browns ≤ 48¢
- Jets at Lions — Model winner: Lions (70%)
- Texans at Colts — Model winner: Colts (52%) · The play: Colts ≤ 45¢
- Patriots at Jaguars — Model winner: Jaguars (50%)
- Chiefs at Dolphins — Model winner: Dolphins (50%) · The play: Dolphins ≤ 42¢
- Titans at Giants — Model winner: Giants (65%) · The play: Giants ≤ 60¢
- Bengals at Steelers — Model winner: Steelers (58%) · The play: Steelers ≤ 48¢
- Seahawks at Commanders — Model winner: Seahawks (63%)
- Cardinals at 49ers — Model winner: 49ers (65%) · The play: Cardinals ≤ 31¢
- Vikings at Buccaneers — Model winner: Buccaneers (64%)
- Ravens at Cowboys — Model winner: Cowboys (52%) · The play: Cowboys ≤ 45¢
- Raiders at Saints — Model winner: Saints (67%)
- Rams at Broncos — Model winner: Rams (54%)
- Eagles at Bears — Model winner: Bears (56%)
All 16 Week 3 games with our DAEPA model win probabilities, the Kalshi price, and our pick on every game. Tap any game for the full breakdown, the Kalshi line and the edge.
Go Pro · 26% off your first year
Every NFL line that moves before kickoff — before the market catches up
Pro is the layer around the free game edges: model win probability against the live Kalshi price on every market, edge alerts in Discord the moment a gap opens, and the full win-total, championship and projection boards.
$111 for the first year, then $150/yr. Code applies automatically. Ends Dec 31. Monthly is $14.99/mo.
Every signal graded in public, losers left on the board. See the record →
Week 3 straight-up picks
6 market favorites and 10 upsets across 16 games. Biggest upset call: Saints over Raiders at 67% (Kalshi 48¢).
- Packers over FalconsUpset66%48¢
- Bills over Chargers65%59.5¢
- Browns over PanthersUpset55%47.5¢
- Lions over Jets70%68.5¢
- Colts over TexansUpset52%45¢
- Jaguars over Patriots50%54.5¢
- Dolphins over ChiefsUpset50%41.5¢
- Giants over Titans65%59.5¢
- Steelers over BengalsUpset58%48¢
- Seahawks over Commanders63%53.5¢
- 49ers over Cardinals65%72.5¢
- Buccaneers over VikingsUpset64%41.5¢
- Cowboys over RavensUpset52%44.5¢
- Saints over RaidersUpset67%48¢
- Rams over BroncosUpset54%48¢
- Bears over EaglesUpset56%42.5¢
Winner = the side with the higher DAEPA model win probability. Kalshi = the mid price of that side’s contract, in cents. An upset is a model winner priced under 50¢; that is a straight-up call, not the play — the play on each game is the edge side in the board above.
Week 3 projected scores
- FalconsatPackers21.4–27.448.8
- ChargersatBills20.3–27.748.0
- PanthersatBrowns18.8–20.339.1
- JetsatLions17.2–26.443.5
- TexansatColts21.2–21.843.0
- PatriotsatJaguars24.2–24.348.4
- ChiefsatDolphins22.9–23.045.9
- TitansatGiants20.3–25.345.7
- BengalsatSteelers22.0–24.646.7
- SeahawksatCommanders26.2–22.048.2
- Cardinalsat49ers22.5–29.952.4
- VikingsatBuccaneers19.6–24.043.6
- RavensatCowboys26.2–26.752.9
- RaidersatSaints17.2–23.640.7
- RamsatBroncos24.6–23.247.8
- EaglesatBears21.9–23.845.8
Projected score = the model margin split around a ratings-implied total (league baseline 44 points plus both offensive ratings, minus both defensive ratings). The total is a derivation from the power ratings, not a stored engine number; the margin is the same one behind the model line above. One decimal, so a projection is never mistaken for a final score.
MIN
-2.6
TB
-0.6
LV
-6.0
NO
-2.3
ATL
-0.4
GB
+3.0
NFL Week 3 predictions — frequently asked
Who does the model pick in NFL Week 3 2026?
Our DAEPA model disagrees with the Kalshi price by 5 points or more on 8 of 16 Week 3 games. The biggest gap is Bengals at Steelers, where the pick is the Steelers at a 58% model win probability against a Kalshi price of 48¢. Every game's pick, model line and edge is in the board above.
How are these NFL Week 3 predictions made?
Every game runs through DAEPA (Defense-Adjusted Expected Points Added), the play-by-play engine behind our PWR power ratings. It produces a neutral-field margin and a home win probability for each matchup, then the pick is wherever that probability and the Kalshi price disagree. Backtested on 3,266 games from 2013 to 2024, DAEPA picks the straight-up winner 59.6% of the time.
What does the edge on each game mean?
Edge is the model win probability minus the Kalshi price for the same side, in percentage points. A +8pp edge means the model has a team at 62% while the market prices its contract at 54¢. We only call a pick when the gap is 5 points or more; anything smaller sits inside the noise of both numbers, so the row shows no edge.
Who wins straight up in NFL Week 3 2026?
The model calls 6 market favorites and 10 upsets straight up in Week 3. Packers over Falcons (66%); Bills over Chargers (65%); Browns over Panthers (55%); Lions over Jets (70%); Colts over Texans (52%); Jaguars over Patriots (50%); Dolphins over Chiefs (50%); Giants over Titans (65%); Steelers over Bengals (58%); Seahawks over Commanders (63%); 49ers over Cardinals (65%); Buccaneers over Vikings (64%); Cowboys over Ravens (52%); Saints over Raiders (67%); Rams over Broncos (54%); Bears over Eagles (56%). The winner is the side with the higher DAEPA model win probability; an upset is a model winner priced under 50¢ on Kalshi. The straight-up winner and the play are different calls — the play is the side where the model and the price disagree by 5 points or more.
What are the projected scores for NFL Week 3?
Projected scores, winner first: Packers 27.4, Falcons 21.4; Bills 27.7, Chargers 20.3; Browns 20.3, Panthers 18.8; Lions 26.4, Jets 17.2; Colts 21.8, Texans 21.2; Jaguars 24.3, Patriots 24.2; Dolphins 23.0, Chiefs 22.9; Giants 25.3, Titans 20.3; Steelers 24.6, Bengals 22.0; Seahawks 26.2, Commanders 22.0; 49ers 29.9, Cardinals 22.5; Buccaneers 24.0, Vikings 19.6; Cowboys 26.7, Ravens 26.2; Saints 23.6, Raiders 17.2; Rams 24.6, Broncos 23.2; Bears 23.8, Eagles 21.9. Each score splits the model margin around a ratings-implied total (league baseline 44 points plus both offensive ratings, minus both defensive ratings) — the total is derived from the power ratings, not a stored engine number, which is why scores carry one decimal.
How often are the Week 3 picks updated?
Daily through kickoff. Kalshi prices are re-captured and the model reruns against them, so the Updated stamp under the title is the last time a number on this page actually moved; it is never a render clock. Power ratings refresh weekly once games are played.
Why read this pick instead of a preview from a big sports outlet?
Two things a bigger outlet does not do. The number is posted before kickoff against the live Kalshi price, so you can see where the model and the market disagree while it still matters. And the page keeps the call after the game — result, hit or miss, closing-line value — so the record is public. Every graded call is at predictionmarketspicks.com/track-record.
Model win probabilities from the DAEPA power-rating engine. Prediction markets carry risk; contracts can settle at $0. Informational only — trade responsibly.