NFL Betting Model

Opponent-adjusted NFL predictions vs. the market. Personal research project, not financial advice.

5
Picks Tracked
3/5
Moneyline (Straight-Up)
2-2-1
Against the Spread
9.4 pts§
Avg. Margin Error
$480.27
Paper Bankroll

Weekly Performance

WeekMoneylineSpread
2026 Week 11-1 (50%)0-1-1
2025 Super Bowl0-1 (0%)0-1-0
2025 Conference Championship1-0 (100%)1-0-0
2025 Divisional1-0 (100%)1-0-0

Weekly Trends

Moneyline Win %

100%
100%
2025 Divisional
100%
100%
2025 Conference Championship
0%
100%
2025 Super Bowl
50%
50%
2026 Week 1
Model moneyline win%Favorite baseline (same week)

Against the Spread Win %

Breakeven 52.4%
100%
2025 Divisional
100%
2025 Conference Championship
0%
2025 Super Bowl
0%
2026 Week 1
ATS win% Breakeven at -110 (52.4%)

Margin Accuracy: Model vs. Market

1.8 vs 9.0
2025 Divisional
0.9 vs 1.5
2025 Conference Championship
15.3 vs 11.5
2025 Super Bowl
14.5 vs 11.8
2026 Week 1
Model avg error (pts)Market avg error (pts) · lower is better

§ Margin error for one game is |predicted margin − actual final margin|, in points -- e.g. picking a team to win by 10 when they win by 14 is a 4-point error. Averaging that across every graded game gives the "Avg. Margin Error" stat above and each week's bars in this chart. Lower is better for both the model's own number and the market's -- the real question this chart tracks is whether the model is closing the gap on the market's own accuracy over time, not just whether picks are winning.

Win Probability Accuracy: Model vs. Polymarket¶

0.280 vs 0.278
2026 Week 1 (2)
Model Brier scorePolymarket Brier score · lower is better

¶ A post-game accuracy check only -- Polymarket's pre-game win probability for each game is pulled and stored before kickoff, then graded afterward the same way the model's own is, using Brier score (squared error between the probability and the actual 0/1 outcome -- e.g. predicting 70% and winning scores (0.7−1)²=0.09, losing scores (0.7−0)²=0.49; lower is better). Nothing here feeds into the moneyline or spread picks above -- those are generated independently from the model's own features, same as always. The number in parentheses next to each week is how many of that week's games actually had a matching Polymarket market to grade against.

Picks

CHI @ CAR
Sun Sep 13, 1:00 PM EDT
Model: CAR by -2.6 (39% win prob) · Market: CHI by 3.0
Moneyline pick: CHI (61% win prob) MEDIUM Confidence†
Spread pick: CAR +3.0 (-108) LOW Confidence†
Low-confidence lean on CAR (+3 at home) -- Chicago still profiles as the better team by ELO and recent efficiency, so this is a thin value play on the number rather than a pick of the better team.
  • Chicago carries a 96-point ELO advantage over Carolina (1493 vs 1397), and pre-game ELO alone gives Carolina just a 42% win probability.
  • Chicago has also been more efficient on the ground recently (+0.007 rushing EPA/play vs -0.182) and has a higher explosive-play rate (13% vs 11%) over the last 8 games.
  • Chicago has played at a faster pace too (27.0s vs 29.2s per play), but the model still sees Carolina as live to cover this small a number.
Per-model breakdown (home win probability)
Logistic Regression30%
Random Forest36%
XGBoost30%
LightGBM40%
Extra Trees39%
TB @ CIN
Sun Sep 13, 1:00 PM EDT
Model: CIN by +3.8 (53% win prob) · Market: CIN by 3.5
Moneyline pick: CIN (53% win prob) LOW Confidence†
Spread pick: CIN -3.5 (-110) LOW Confidence†
Slim, low-confidence lean on CIN to cover as home favorites -- Tampa Bay actually carries the larger ELO edge, but Cincinnati's better recent offensive success rate and pass rush give it a marginal case to justify the number.
  • Tampa Bay holds a 27-point ELO advantage (1518 vs 1491), the largest single factor working against Cincinnati here.
  • Cincinnati has still been more efficient recently: a better offensive success rate (47% vs 44%) and pass-rush pressure rate (17% vs 12%) over the last 8 games.
  • Cincinnati also has the rushing efficiency edge (+0.007 EPA/play vs -0.064) and has covered the spread in exactly half of its last 8 games.
Per-model breakdown (home win probability)
Logistic Regression52%
Random Forest53%
XGBoost54%
LightGBM43%
Extra Trees57%
NO @ DET
Sun Sep 13, 1:00 PM EDT
Model: DET by +5.4 (61% win prob) · Market: DET by 7.0
Moneyline pick: DET (61% win prob) MEDIUM Confidence†
Spread pick: NO +7.0 (-110) LOW Confidence†
Recommended wager: $6.07 (1.3% of $480.27 paper bankroll, 25% Kelly) · cover probability 55%
Low-confidence lean on NO to cover as big underdogs -- Detroit is still the clearly better team by ELO, but New Orleans' modest edges in rushing efficiency and explosive-plays-allowed are enough for the model to think the current 7-point number is too large.
  • Detroit carries a 137-point ELO advantage (1579 vs 1442) and a pre-game win probability of 74%, so this remains a clear favorite-underdog matchup.
  • New Orleans has had the better rushing efficiency recently (-0.116 EPA/play vs -0.134) and has allowed fewer explosive plays (10% vs 13%) over the last 8 games.
  • Detroit still holds the explosive-play-rate edge on offense (15% vs 12%), but not enough, in the model's view, to fully justify a full touchdown spread.
Per-model breakdown (home win probability)
Logistic Regression62%
Random Forest61%
XGBoost66%
LightGBM77%
Extra Trees59%
BUF @ HOU
Sun Sep 13, 2:00 PM EDT
Model: HOU by +6.4 (48% win prob) · Market: BUF by 1.5
Moneyline pick: BUF (52% win prob) LOW Confidence†
Spread pick: HOU +1.5 (-110) HIGH Confidence†
Recommended wager: $51.57 (10.7% of $480.27 paper bankroll, 25% Kelly) · cover probability 73%
High-confidence pick on HOU to cover as home underdogs -- the market only favors Buffalo by 1.5 points, and Houston's notably faster recent pace is enough for the model to back the home dog covering despite Buffalo's broader statistical edge.
  • Buffalo enters with more of the traditional efficiency edges: a 61-point ELO advantage (1626 vs 1565), the better recent offensive success rate (50% vs 41%), and edges in rushing (+0.070 vs -0.167 EPA/play) and early-down (+0.097 vs -0.119) efficiency.
  • Houston's standout data point is tempo -- the Texans have played at a notably faster pace recently (27.5s vs 31.5s per snap).
  • With the market favoring Buffalo by only 1.5 points, the model sees enough value to back Houston to cover as the home underdog.
Per-model breakdown (home win probability)
Logistic Regression47%
Random Forest42%
XGBoost54%
LightGBM59%
Extra Trees44%
BAL @ IND
Sun Sep 13, 1:00 PM EDT
Model: IND by -6.6 (37% win prob) · Market: BAL by 3.5
Moneyline pick: BAL (63% win prob) MEDIUM Confidence†
Spread pick: BAL -3.5 (-110) MEDIUM Confidence†
Recommended wager: $17.70 (3.7% of $480.27 paper bankroll, 25% Kelly) · cover probability 59%
Medium-confidence pick on BAL to cover as road favorites -- the Ravens carry a big statistical edge in ELO and rushing efficiency, and the model thinks the market's 3.5-point number undersells how lopsided this one is.
  • Baltimore holds a 102-point ELO advantage (1577 vs 1475), and pre-game ELO gives Indianapolis just a 41% win probability.
  • Baltimore has also been much more efficient on the ground recently (+0.111 EPA/play vs -0.116) over the last 8 games.
  • Indianapolis' case rests mostly on red-zone efficiency -- a 69% touchdown rate compared to Baltimore's 40% -- and a more pass-heavy approach overall (60% vs 50%).
Per-model breakdown (home win probability)
Logistic Regression26%
Random Forest37%
XGBoost33%
LightGBM31%
Extra Trees38%
CLE @ JAX
Sun Sep 13, 1:00 PM EDT
Model: JAX by -2.4 (70% win prob) · Market: JAX by 8.5
Moneyline pick: JAX (70% win prob) MEDIUM Confidence†
Spread pick: CLE +8.5 (-110) HIGH Confidence†
Recommended wager: $69.40 (14.5% of $480.27 paper bankroll, 25% Kelly) · cover probability 80%
High-confidence lean on CLE to cover as big road underdogs -- Jacksonville still profiles as the better team across the board, but the model doesn't think a 71%-win-probability team deserves a full 8.5-point line, making the Browns getting all those points the value side.
  • Jacksonville has the edge in nearly every underlying measure: pre-game ELO gives the Jaguars a 71% win probability, backed by a 112-point ELO advantage (1524 vs 1413).
  • Jacksonville has also been more efficient offensively (+0.080 EPA/play vs -0.231, a 45% vs 35% success rate) and has leaned into a more pass-heavy approach in neutral game scripts (64% vs 47%).
  • None of that is disputed -- the model's disagreement with the market is about magnitude, not direction, and it isn't willing to lay a full 8.5 points on it.
Per-model breakdown (home win probability)
Logistic Regression78%
Random Forest75%
XGBoost69%
LightGBM69%
Extra Trees73%
ATL @ PIT
Sun Sep 13, 1:00 PM EDT
Model: PIT by +1.0 (54% win prob) · Market: PIT by 6.0
Moneyline pick: PIT (54% win prob) LOW Confidence†
Spread pick: ATL +6.0 (-110) MEDIUM Confidence†
Recommended wager: $31.53 (6.6% of $480.27 paper bankroll, 25% Kelly) · cover probability 65%
Medium-confidence lean on ATL to cover as road underdogs -- Pittsburgh is still the stronger team by ELO, but Atlanta's edges in offensive success rate and explosive plays make the model think a 6-point line is more than the Steelers deserve.
  • Pittsburgh holds the ELO edge (a 60-point advantage, 1508 vs 1447) and a 64% pre-game win probability.
  • Atlanta has the better recent offensive success rate (44% vs 42%) and explosive-play rate (13% vs 11%) over the last 8 games.
  • Pittsburgh's defense has been marginally better at limiting success rate (42% allowed vs 43%), but not enough, in the model's view, to justify laying a full 6 points.
Per-model breakdown (home win probability)
Logistic Regression46%
Random Forest57%
XGBoost64%
LightGBM67%
Extra Trees52%
NYJ @ TEN
Sun Sep 13, 2:00 PM EDT
Model: TEN by -1.3 (48% win prob) · Market: TEN by 1.5
Moneyline pick: NYJ (52% win prob) LOW Confidence†
Spread pick: NYJ +1.5 (-110) LOW Confidence†
Recommended wager: $15.53 (3.2% of $480.27 paper bankroll, 25% Kelly) · cover probability 59%
Low-confidence lean on NYJ to cover as home underdogs -- Tennessee has the sharper pass rush and explosive-play numbers, but the Titans have covered just 12% of their last 8 spreads, and the model doesn't think a short favorite's role changes that.
  • Tennessee has the edge in recent pass-rush pressure (15% vs 9%) and explosive-play rate (9% vs 7%) over the last 8 games.
  • The Jets have played at a faster pace recently (27.1s vs 28.9s per play).
  • Tennessee has covered the spread in just 12% of its last 8 games, a cold recent trend working against them as favorites here.
Per-model breakdown (home win probability)
Logistic Regression40%
Random Forest52%
XGBoost36%
LightGBM35%
Extra Trees53%
ARI @ LAC
Sun Sep 13, 7:25 PM EDT
Model: LAC by -2.9 (53% win prob) · Market: LAC by 9.5
Moneyline pick: LAC (53% win prob) LOW Confidence†
Spread pick: ARI +9.5 (-110) HIGH Confidence†
Recommended wager: $77.23 (16.1% of $480.27 paper bankroll, 25% Kelly) · cover probability 83%
High-confidence lean on ARI to cover as big road underdogs -- the Chargers are still the stronger team by ELO, but Arizona's edges in offensive efficiency and explosive plays make a 9.5-point number look too steep to the model.
  • The Chargers carry a 70% pre-game win probability by ELO, the largest factor working against Arizona.
  • Arizona has been the more efficient offense recently, with better EPA per play (-0.031 vs -0.119) and a higher explosive-play rate (13% vs 9%) over the last 8 games.
  • Arizona has also leaned heavily pass-first, both in neutral game scripts (75% vs 59%) and overall (74% vs 63%).
Per-model breakdown (home win probability)
Logistic Regression56%
Random Forest54%
XGBoost41%
LightGBM45%
Extra Trees53%
MIA @ LV
Sun Sep 13, 7:25 PM EDT
Model: LV by -7.8 (40% win prob) · Market: LV by 3.0
Moneyline pick: MIA (60% win prob) LOW Confidence†
Spread pick: MIA +3.0 (-105) HIGH Confidence†
Recommended wager: $70.09 (14.6% of $480.27 paper bankroll, 25% Kelly) · cover probability 80%
High-confidence pick on MIA to cover as road underdogs -- Miami actually rates as the better team by ELO despite Las Vegas being favored by the market, with a 98-point ELO edge and better rushing efficiency giving the model real conviction here.
  • Miami carries a 98-point ELO advantage over Las Vegas (1464 vs 1367), and pre-game ELO gives the Raiders just a 42% win probability.
  • Miami has also had much better rushing efficiency recently (-0.043 EPA/play vs -0.308) over the last 8 games.
  • Las Vegas has played at a faster pace (28.7s vs 31.4s per play) and leaned more pass-heavy in neutral scripts (62% vs 46%), but the model sees Miami as the sounder team here despite being priced as a home underdog.
Per-model breakdown (home win probability)
Logistic Regression23%
Random Forest43%
XGBoost40%
LightGBM58%
Extra Trees37%
GB @ MIN
Sun Sep 13, 5:25 PM EDT
Model: MIN by +10.3 (53% win prob) · Market: MIN by 1.5
Moneyline pick: MIN (53% win prob) LOW Confidence†
Spread pick: MIN -1.5 (-110) HIGH Confidence†
Recommended wager: $57.24 (11.9% of $480.27 paper bankroll, 25% Kelly) · cover probability 75%
High-confidence pick on MIN to cover at home -- Green Bay has the sharper passing numbers, but Minnesota's rushing efficiency and its defense's knack for limiting explosive plays give the model real conviction that a 1.5-point line badly undersells the Vikings.
  • Green Bay has the passing efficiency edge (+0.232 EPA/play vs -0.204) and a higher explosive-play rate (14% vs 11%) over the last 8 games.
  • Minnesota has been the better rushing team (-0.017 EPA/play vs -0.189) and its defense has allowed fewer explosive plays (9% vs 13%).
  • Minnesota's broader offensive efficiency edge (+0.090 EPA/play vs -0.134) rounds out the case for a bigger margin than the market's 1.5-point line implies.
Per-model breakdown (home win probability)
Logistic Regression49%
Random Forest52%
XGBoost55%
LightGBM61%
Extra Trees52%
WAS @ PHI
Sun Sep 13, 4:25 PM EDT
Model: PHI by +6.2 (63% win prob) · Market: PHI by 6.0
Moneyline pick: PHI (63% win prob) MEDIUM Confidence†
Spread pick: PHI -6.0 (-110) LOW Confidence†
Medium-confidence pick on PHI to cover at home -- the Eagles hold a commanding edge across ELO, pass-heavy game script, and defensive efficiency, though the model's edge over the market number itself is thin.
  • Philadelphia carries a 149-point ELO advantage over Washington (1602 vs 1453) and a 75% pre-game win probability.
  • Philadelphia's defense has been excellent, with a better EPA allowed per play (-0.066 vs +0.184) and a lower explosive-play rate allowed (9% vs 15%) over the last 8 games.
  • Philadelphia has also leaned pass-heavy in neutral game scripts (65% vs 48%).
Per-model breakdown (home win probability)
Logistic Regression53%
Random Forest65%
XGBoost73%
LightGBM85%
Extra Trees66%
DAL @ NYG
Sun Sep 13, 8:20 PM EDT
Model: NYG by -4.1 (44% win prob) · Market: DAL by 3.0
Moneyline pick: DAL (56% win prob) LOW Confidence†
Spread pick: DAL -3.0 (-105) LOW Confidence†
Recommended wager: $5.64 (1.2% of $480.27 paper bankroll, 25% Kelly) · cover probability 54%
Low-confidence lean on DAL to cover as road favorites -- the Cowboys hold the edges in ELO, red-zone efficiency, and recent head-to-head scoring margin, even if the number itself is close to a coinflip in the model's eyes.
  • Dallas carries a 65-point ELO advantage (1473 vs 1408) and the better red-zone touchdown rate (49% vs 40%) over the last 8 games.
  • Over their last 6 meetings, Dallas has outscored the Giants by an average of 11.7 points.
  • New York's lone edge is rushing efficiency (-0.024 EPA/play vs -0.037), not enough for the model to move off Dallas here.
Per-model breakdown (home win probability)
Logistic Regression41%
Random Forest41%
XGBoost39%
LightGBM50%
Extra Trees44%
DEN @ KC
Mon Sep 14, 9:15 PM EDT
Model: KC by -0.7 (46% win prob) · Market: KC by 2.5
Moneyline pick: DEN (54% win prob) LOW Confidence†
Spread pick: DEN +2.5 (-108) MEDIUM Confidence†
Recommended wager: $19.63 (4.1% of $480.27 paper bankroll, 25% Kelly) · cover probability 60%
Medium-confidence lean on DEN to cover as road underdogs -- Denver has been the more efficient offense recently by success rate and EPA per play, and the model doesn't think Kansas City's 2.5-point favorite tag is fully earned.
  • Denver has the better recent offensive success rate (44% vs 38%) and offensive EPA per play (+0.014 vs -0.121) over the last 8 games.
  • Denver also has a slight edge in explosive-play rate (10% vs 8%).
  • Kansas City's advantages are narrower -- a faster recent pace (27.1s vs 29.0s per play) and a marginal rushing-efficiency edge (-0.114 vs -0.125 EPA/play).
Per-model breakdown (home win probability)
Logistic Regression43%
Random Forest41%
XGBoost45%
LightGBM50%
Extra Trees47%
DET @ BUF
Thu Sep 17, 8:15 PM EDT
Model: BUF by +9.6 (69% win prob) · Market: BUF by 3.0
Moneyline pick: BUF (69% win prob) MEDIUM Confidence†
Spread pick: BUF -3.0 (-112) MEDIUM Confidence†
Recommended wager: $41.88 (8.7% of $480.27 paper bankroll, 25% Kelly) · cover probability 69%
Medium-confidence pick on BUF to cover at home -- Buffalo's edges in offensive EPA and rushing efficiency give the model more conviction than the market's 3-point line reflects.
  • Buffalo carries a 47-point ELO advantage over Detroit (1626 vs 1579) and a 62% pre-game win probability.
  • Buffalo has been the more efficient offense recently, in both overall EPA per play (+0.162 vs -0.004) and rushing EPA per play (+0.070 vs -0.134).
  • Buffalo is also playing under a new head coach this season, a contextual factor the model flags alongside the statistical edges.
Per-model breakdown (home win probability)
Logistic Regression69%
Random Forest76%
XGBoost77%
LightGBM82%
Extra Trees73%
CAR @ ATL
Sun Sep 20, 1:00 PM EDT
Model: ATL by +4.5 (65% win prob) · Market: ATL by 1.5
Moneyline pick: ATL (65% win prob) MEDIUM Confidence†
Spread pick: ATL -1.5 (-110) MEDIUM Confidence†
Recommended wager: $17.13 (3.6% of $480.27 paper bankroll, 25% Kelly) · cover probability 59%
Medium-confidence pick on ATL to cover at home -- the Falcons hold the ELO and rushing-efficiency edges, and the model thinks a 1.5-point line undersells how much better they've projected recently.
  • Atlanta carries a 51-point ELO advantage over Carolina (1447 vs 1397) and a 63% pre-game win probability.
  • Atlanta has also had the better rushing efficiency recently (-0.092 EPA/play vs -0.182) and a faster pace (28.4s vs 29.2s per play).
  • Carolina's case rests mainly on leaning more pass-heavy overall (65% vs 53%), which hasn't been enough to close the gap in the model's view.
Per-model breakdown (home win probability)
Logistic Regression67%
Random Forest67%
XGBoost73%
LightGBM79%
Extra Trees66%
NO @ BAL
Sun Sep 20, 1:00 PM EDT
Model: BAL by +5.3 (67% win prob) · Market: BAL by 7.5
Moneyline pick: BAL (67% win prob) MEDIUM Confidence†
Spread pick: NO +7.5 (-110) LOW Confidence†
Recommended wager: $10.94 (2.3% of $480.27 paper bankroll, 25% Kelly) · cover probability 57%
Low-confidence lean on NO to cover as big road underdogs -- Baltimore is clearly the better team by ELO and rushing efficiency, but the model doesn't think that translates into a full 7.5-point margin.
  • Baltimore carries a 136-point ELO advantage over New Orleans (1577 vs 1442) and a 73% pre-game win probability.
  • Baltimore has also had the better rushing efficiency (+0.111 EPA/play vs -0.116) and explosive-play rate (14% vs 12%) over the last 8 games.
  • New Orleans' one edge is a faster recent pace (27.5s vs 29.7s per play), enough for the model to see value in the Saints simply covering a large number.
Per-model breakdown (home win probability)
Logistic Regression69%
Random Forest69%
XGBoost81%
LightGBM80%
Extra Trees66%
MIN @ CHI
Sun Sep 20, 2:00 PM EDT
Model: CHI by +2.0 (56% win prob) · Market: CHI by 3.0
Moneyline pick: CHI (56% win prob) LOW Confidence†
Spread pick: MIN +3.0 (-110) LOW Confidence†
Recommended wager: $2.09 (0.4% of $480.27 paper bankroll, 25% Kelly) · cover probability 53%
Low-confidence lean on MIN to cover as road underdogs -- the Vikings hold the ELO and pass-rush edges, and the model doesn't think Chicago's recent ATS success fully offsets that.
  • Minnesota carries a 47-point ELO advantage over Chicago (1539 vs 1493) and the better pass-rush pressure rate (21% vs 13%) over the last 8 games.
  • Minnesota's defense has also allowed a lower success rate (40% vs 45%).
  • Chicago has played at a faster pace (27.0s vs 30.4s per play) and has covered the spread in 62% of its last 8 games, but the model still gives Minnesota the edge here.
Per-model breakdown (home win probability)
Logistic Regression48%
Random Forest61%
XGBoost55%
LightGBM48%
Extra Trees64%
CIN @ HOU
Sun Sep 20, 2:00 PM EDT
Model: HOU by +14.6 (61% win prob) · Market: HOU by 2.5
Moneyline pick: HOU (61% win prob) MEDIUM Confidence†
Spread pick: HOU -2.5 (-110) HIGH Confidence†
Recommended wager: $75.76 (15.8% of $480.27 paper bankroll, 25% Kelly) · cover probability 82%
High-confidence pick on HOU to cover at home -- the Texans' ELO edge and faster recent pace give the model real conviction that a 2.5-point favorite tag is far too modest here.
  • Houston carries a 75-point ELO advantage over Cincinnati (1565 vs 1491) and a 66% pre-game win probability.
  • Houston has also played at a faster pace recently (27.5s vs 29.3s per play).
  • Cincinnati's case rests on better rushing efficiency (+0.007 EPA/play vs -0.167) and offensive success rate (47% vs 41%), but the model sees Houston as a far bigger favorite than the market number suggests.
Per-model breakdown (home win probability)
Logistic Regression64%
Random Forest61%
XGBoost74%
LightGBM72%
Extra Trees59%
PIT @ NE
Sun Sep 20, 1:00 PM EDT
Model: NE by +7.2 (65% win prob) · Market: NE by 4.5
Moneyline pick: NE (65% win prob) MEDIUM Confidence†
Spread pick: NE -4.5 (-108) LOW Confidence†
Recommended wager: $15.88 (3.3% of $480.27 paper bankroll, 25% Kelly) · cover probability 58%
Low-confidence pick on NE to cover at home -- the Patriots hold a modest ELO edge, though Pittsburgh actually rates as the better team once strength of schedule is factored in, keeping the model's conviction limited.
  • New England carries a 25-point ELO advantage over Pittsburgh (1533 vs 1508) and the better explosive-play rate (13% vs 11%) over the last 8 games.
  • Pittsburgh has actually rated as the better team once opponent strength is factored in, outrating New England by 3.1 points of opponent-adjusted margin (+0.1 vs -3.0).
  • New England has the edges in pass-rush pressure (21% vs 12%), while Pittsburgh's lone efficiency edge is rushing (-0.019 EPA/play vs -0.131).
Per-model breakdown (home win probability)
Logistic Regression63%
Random Forest68%
XGBoost78%
LightGBM73%
Extra Trees68%
GB @ NYJ
Sun Sep 20, 1:00 PM EDT
Model: NYJ by -7.0 (31% win prob) · Market: GB by 5.5
Moneyline pick: GB (69% win prob) MEDIUM Confidence†
Spread pick: GB -5.5 (-108) LOW Confidence†
Recommended wager: $6.66 (1.4% of $480.27 paper bankroll, 25% Kelly) · cover probability 55%
Low-confidence pick on GB to cover as road favorites -- the Packers hold a lopsided edge across ELO, passing efficiency, and explosiveness, even if the projected margin is close enough to the market number to keep confidence modest.
  • Green Bay carries a 188-point ELO advantage over the Jets (1554 vs 1367), and pre-game ELO gives New York just a 30% win probability.
  • Green Bay has the passing efficiency edge (+0.232 EPA/play vs -0.220) and a much higher explosive-play rate (14% vs 7%) over the last 8 games.
  • Green Bay's offensive success rate is also well ahead (47% vs 38%).
Per-model breakdown (home win probability)
Logistic Regression20%
Random Forest27%
XGBoost15%
LightGBM14%
Extra Trees33%
CLE @ TB
Sun Sep 20, 1:00 PM EDT
Model: TB by -5.2 (63% win prob) · Market: TB by 6.5
Moneyline pick: TB (63% win prob) MEDIUM Confidence†
Spread pick: CLE +6.5 (-110) HIGH Confidence†
Recommended wager: $73.28 (15.3% of $480.27 paper bankroll, 25% Kelly) · cover probability 81%
High-confidence lean on CLE to cover as big road underdogs -- Tampa Bay is clearly the better team across ELO, red-zone efficiency, and passing, but the model doesn't think that gap is worth a full 6.5 points.
  • Tampa Bay carries a 105-point ELO advantage over Cleveland (1518 vs 1413) and a 70% pre-game win probability.
  • Tampa Bay has also been much sharper in the red zone (52% touchdown rate vs 29%) and through the air (+0.008 EPA/play vs -0.253) over the last 8 games.
  • Tampa Bay has covered the spread in 62% of its last 8 games, but the model still sees Cleveland as live to cover this large a number.
Per-model breakdown (home win probability)
Logistic Regression74%
Random Forest64%
XGBoost64%
LightGBM42%
Extra Trees67%
PHI @ TEN
Sun Sep 20, 2:00 PM EDT
Model: TEN by -11.4 (32% win prob) · Market: PHI by 5.5
Moneyline pick: PHI (68% win prob) MEDIUM Confidence†
Spread pick: PHI -5.5 (-110) MEDIUM Confidence†
Recommended wager: $38.02 (7.9% of $480.27 paper bankroll, 25% Kelly) · cover probability 67%
Medium-confidence pick on PHI to cover on the road -- a massive 244-point ELO gap and edges across explosiveness, success rate, and rushing efficiency have the model projecting a much bigger Eagles win than the market's 5.5-point line.
  • Philadelphia carries a 244-point ELO advantage over Tennessee (1602 vs 1358), and pre-game ELO gives the Titans just a 24% win probability.
  • Philadelphia has the edges in offensive success rate (43% vs 37%), explosive-play rate (13% vs 9%), and rushing efficiency (-0.040 EPA/play vs -0.093) over the last 8 games.
Per-model breakdown (home win probability)
Logistic Regression19%
Random Forest29%
XGBoost15%
LightGBM13%
Extra Trees35%
JAX @ DEN
Sun Sep 20, 6:05 PM EDT
Model: DEN by +5.1 (54% win prob) · Market: DEN by 3.0
Moneyline pick: DEN (54% win prob) LOW Confidence†
Spread pick: DEN -3.0 (-108) LOW Confidence†
Recommended wager: $11.39 (2.4% of $480.27 paper bankroll, 25% Kelly) · cover probability 56%
Low-confidence pick on DEN to cover at home -- Denver holds the ELO edge, but Jacksonville's better opponent-adjusted rating and offensive efficiency keep the model's confidence limited here.
  • Denver carries a 50-point ELO advantage over Jacksonville (1574 vs 1524) and a 63% pre-game win probability.
  • Jacksonville has actually rated as the better team once strength of schedule is factored in, outrating Denver by 2.6 points of opponent-adjusted margin (+4.5 vs +1.8).
  • Jacksonville also has the edges in offensive EPA per play (+0.080 vs +0.014) and explosive-play rate (15% vs 10%).
Per-model breakdown (home win probability)
Logistic Regression51%
Random Forest55%
XGBoost62%
LightGBM64%
Extra Trees52%
LV @ LAC
Sun Sep 20, 7:05 PM EDT
Model: LAC by +1.2 (62% win prob) · Market: LAC by 9.5
Moneyline pick: LAC (62% win prob) MEDIUM Confidence†
Spread pick: LV +9.5 (-112) HIGH Confidence†
Recommended wager: $53.39 (11.1% of $480.27 paper bankroll, 25% Kelly) · cover probability 74%
High-confidence lean on LV to cover as big road underdogs -- the Chargers are clearly the better team by ELO and opponent-adjusted rating, but the model doesn't think that gap is worth a full 9.5 points, especially with LA having covered just 38% of its recent spreads.
  • The Chargers carry a 150-point ELO advantage over Las Vegas (1517 vs 1367) and a 75% pre-game win probability.
  • The Chargers have also rated as the better team once strength of schedule is factored in, outrating Las Vegas by 5.7 points of opponent-adjusted margin (+0.5 vs -5.1).
  • Despite that, LA has covered the spread in just 38% of its last 8 games and is playing under a new head coach this season -- the model doesn't think this is a 9.5-point favorite.
Per-model breakdown (home win probability)
Logistic Regression75%
Random Forest62%
XGBoost63%
LightGBM49%
Extra Trees64%
SEA @ ARI
Sun Sep 20, 7:25 PM EDT
Model: ARI by -13.6 (32% win prob) · Market: no line
Moneyline pick: SEA (68% win prob) MEDIUM Confidence†
No market line is posted yet for this one, but on the moneyline the model likes SEA comfortably -- Seattle carries a massive 220-point ELO edge and has allowed far fewer explosive plays than Arizona.
  • Seattle carries a 220-point ELO advantage over Arizona (1629 vs 1409), and pre-game ELO gives the Cardinals just a 26% win probability.
  • Seattle's defense has allowed far fewer explosive plays recently (9% vs 15%) over the last 8 games.
  • Arizona's case is built around pace and pass-heaviness -- a faster snap rate (25.9s vs 30.0s per play) and a heavier pass rate in neutral scripts (75% vs 57%) -- but not enough to threaten the ELO gap.
Per-model breakdown (home win probability)
Logistic Regression17%
Random Forest31%
XGBoost22%
LightGBM31%
Extra Trees29%
WAS @ DAL
Sun Sep 20, 5:25 PM EDT
Model: DAL by -0.0 (52% win prob) · Market: DAL by 4.5
Moneyline pick: DAL (52% win prob) LOW Confidence†
Spread pick: WAS +4.5 (-110) MEDIUM Confidence†
Recommended wager: $28.39 (5.9% of $480.27 paper bankroll, 25% Kelly) · cover probability 64%
Medium-confidence lean on WAS to cover as road underdogs -- Dallas has the recent head-to-head scoring history and a modest ELO edge, but the model doesn't think that's worth a full 4.5 points.
  • Over their last 6 meetings, Dallas has outscored Washington by an average of 16.0 points, and Dallas carries a 20-point ELO advantage (1473 vs 1453).
  • Dallas has also played at a faster pace recently (26.7s vs 29.0s per play) and has the better offensive success rate (47% vs 45%).
  • Washington's one edge is rushing efficiency (-0.028 EPA/play vs -0.037), part of why the model sees the Commanders as live to cover this number.
Per-model breakdown (home win probability)
Logistic Regression45%
Random Forest58%
XGBoost33%
LightGBM49%
Extra Trees55%
MIA @ SF
Sun Sep 20, 7:25 PM EDT
Model: SF by +5.2 (63% win prob) · Market: SF by 11.5
Moneyline pick: SF (63% win prob) MEDIUM Confidence†
Spread pick: MIA +11.5 (-105) MEDIUM Confidence†
Recommended wager: $42.68 (8.9% of $480.27 paper bankroll, 25% Kelly) · cover probability 69%
Medium-confidence lean on MIA to cover as big road underdogs -- San Francisco is the better team by ELO and pace, but an 11.5-point line is more than the model thinks this gap is worth.
  • San Francisco carries a 101-point ELO advantage over Miami (1565 vs 1464) and a 69% pre-game win probability.
  • San Francisco has also played at a faster pace recently (29.0s vs 31.4s per play) and allowed a slightly lower success rate (48% vs 49%).
  • Miami's one edge is a much higher explosive-play rate (16% vs 10%), part of why the model sees value in the Dolphins getting all those points.
Per-model breakdown (home win probability)
Logistic Regression56%
Random Forest67%
XGBoost68%
LightGBM89%
Extra Trees62%
IND @ KC
Sun Sep 20, 9:20 PM EDT
Model: KC by +6.8 (64% win prob) · Market: KC by 6.5
Moneyline pick: KC (64% win prob) MEDIUM Confidence†
Spread pick: KC -6.5 (-112) LOW Confidence†
Low-confidence, razor-thin pick on KC to cover at home -- the Chiefs hold the ELO, pass-rush, and pace edges, but the model's projected margin lines up almost exactly with the market's 6.5 points.
  • Kansas City carries a 67% pre-game win probability by ELO and the better pass-rush pressure rate (18% vs 10%) over the last 8 games.
  • Kansas City has also played at a faster pace (27.1s vs 28.9s per play), with a marginal rushing-efficiency edge (-0.114 vs -0.116 EPA/play).
  • Indianapolis' one edge is a higher explosive-play rate (11% vs 8%), not enough to flip the model off Kansas City.
Per-model breakdown (home win probability)
Logistic Regression67%
Random Forest65%
XGBoost82%
LightGBM79%
Extra Trees61%
NYG @ LA
Mon Sep 21, 11:15 PM EDT
Model: LA by +7.1 (63% win prob) · Market: no line
Moneyline pick: LA (63% win prob) MEDIUM Confidence†
No market line is posted yet for this one, but on the moneyline the model favors LA -- the Rams carry a big ELO edge and the Giants face a cross-country, 3+ time-zone trip, even though NYG's opponent-adjusted rating is oddly the better of the two.
  • The Rams carry a 165-point ELO advantage over the Giants (1574 vs 1408), and pre-game ELO gives Los Angeles a 77% win probability.
  • New York is also traveling west across 3+ time zones for this game.
  • Once strength of schedule is factored in, New York actually rates as the better team, outrating Los Angeles by 18.1 points of opponent-adjusted margin (-1.9 vs -20.0) -- a sign the Rams' underlying performance has lagged their record this season.
Per-model breakdown (home win probability)
Logistic Regression49%
Random Forest66%
XGBoost73%
LightGBM80%
Extra Trees70%