Published on Wed Nov 04 2026 19:30:00 GMT+0000 (Coordinated Universal Time) by cresencio
NFL Week 8 Recap: Ten Winners and a Primetime Spread Sweep
Ten correct winners and three winning primetime spread selections made Week 8 a stronger week for our sides. Totals were another story. The ensemble finished 10–4 on winners, 8–6 against the spread and 5–9 on totals. Green Bay and Jacksonville delivered complete three-market sweeps; Chicago’s Monday upset delivered the spread and over while defeating our Seattle winner pick.
All 14 games are final. We grade the forecasts issued October 28 at 2:30:49 p.m. CDT, at their original lines, as published in the Week 8 preview and combined primetime preview. Below the recap is our first look at all 15 Week 9 markets, ahead of tomorrow’s full preview.
The Week 8 record
| Prediction | Correct | Incorrect | Pushes | Correct rate |
|---|---|---|---|---|
| Outright winner | 10 | 4 | 0 | 71.4% |
| Against the saved spread | 8 | 6 | 0 | 57.1% |
| Saved game total | 5 | 9 | 0 | 35.7% |
The season winner record improves to 73–47, or 60.8%, across 120 decisions. There have been 121 completed games; one tie is excluded from the binary winner record. We retain the original published Week 4 forecasts in that season comparison.
These are records of forecast selections. The prediction board includes games that did not qualify for a betting recommendation, so these percentages cannot be translated into units won or portfolio returns.
Every game against the original picks
Scores read away–home, matching each matchup. The projected scores come from the separate XGBoost margin and total models; winner, spread and total selections come from the probability forecasts. W means correct and L incorrect. LAR denotes the Rams.
| Game | Projected score | Final score | Winner pick | Saved spread pick | Saved total pick |
|---|---|---|---|---|---|
| CAR at GB | 21.7–23.2 | 19–27 | GB / W | GB −3.5 / W | Under 48 / W |
| BAL at BUF | 26.4–30.5 | 10–36 | BUF / W | BAL +3.5 / L | Over 53.5 / L |
| TEN at CIN | 16.5–26.9 | 36–43 | CIN / W | TEN +10.5 / W | Under 46.5 / L |
| ARI at DAL | 22.6–28.3 | 33–30 | DAL / L | DAL −6.5 / L | Over 54.5 / W |
| MIN at DET | 23.0–26.6 | 15–19 | DET / W | DET −3.5 / W | Over 46 / L |
| IND at JAX | 21.8–28.2 | 7–42 | JAX / W | JAX −8 / W | Over 46.5 / W |
| LV at NYJ | 19.8–23.6 | 38–33 | NYJ / L | NYJ −1 / L | Under 45.5 / L |
| CLE at PIT | 20.2–24.3 | 6–27 | PIT / W | CLE +3.5 / L | Over 39.5 / L |
| ATL at TB | 22.5–25.4 | 0–16 | TB / W | TB −2 / W | Over 46 / L |
| LAC at LAR | 19.5–24.9 | 9–24 | LAR / W | LAC +6.5 / L | Over 44.5 / L |
| KC at DEN | 23.1–19.3 | 24–28 | KC / L | KC −1.5 / L | Over 41 / W |
| NE at MIA | 21.7–20.6 | 24–19 | NE / W | MIA +7 / W | Over 44.5 / L |
| PHI at WAS | 17.2–22.9 | 10–27 | WAS / W | WAS −5 / W | Over 40 / L |
| CHI at SEA | 23.9–25.9 | 34–17 | SEA / L | CHI +3 / W | Over 46.5 / W |
Finals and original-line grades come from the verified Week 8 closeout notes. Score projections are rounded independently from their saved full-precision values.
The two strongest outright selections came through: Cincinnati at 75.6% and Jacksonville at 74.3%. Among the four missed winners, Dallas at 63.8% carried the highest confidence; Arizona won 33–30. The Jets, Kansas City and Seattle supplied the other winner misses.
New England again illustrates why a favorite winning does not settle the spread question. The Patriots won 24–19, making both New England to win and Miami +7 correct. Their 43 combined points defeated the saved Over 44.5 selection. Buffalo and Pittsburgh produced the opposite spread outcome: both won comfortably, but our visiting-underdog picks, Baltimore +3.5 and Cleveland +3.5, lost.
Three primetime spread wins, three different score stories
Across the three games in our combined preview, the ensemble went 2–1 on winners, 3–0 on spreads and 2–1 on totals. That is a sweep of the spread selections; only Green Bay swept all three prediction categories.
Thursday: Green Bay 27, Carolina 19. The Packers delivered the 58.9% winner pick, covered −3.5 by 4.5 points and finished with 46 total points, below the original 48. All three selections won. Our published Thursday winner record improves from 4–3 to 5–3.
The point model had projected Green Bay 23.17–Carolina 21.65, a 1.52-point margin and 44.82 total. The actual total was just 1.18 points higher, but the eight-point margin was 6.48 points larger than projected. Getting the under right did not mean we predicted the size of Green Bay’s win. Later quotes, including the Thursday total of 47, do not replace Under 48 in this record.
Sunday night: Washington 27, Philadelphia 10. Washington’s 55.5% winner lean and −5 spread pick both won. The 17-point margin cleared that original handicap by 12 points, but the teams combined for only 37, defeating Over 40.
The separate point forecast was Washington 22.85–Philadelphia 17.20, with a 5.66-point margin and 40.05 total. Washington won by 11.34 more points than the projected margin, while the game produced 3.05 fewer points than the projected total. The favorite won decisively without the game clearing our scoring threshold.
The preview also identified Philadelphia moneyline +230 at FanDuel as the only qualifying primetime recommendation in Thursday’s capture. That quoted selection would have lost with the Eagles’ defeat. It was a separate price-based selection from Washington to win; we have no evidence here of an actual wager or a settled portfolio return.
Monday: Chicago 34, Seattle 17. Seattle was our 56.0% winner pick, and that was wrong. Chicago +3 and Over 46.5 were correct: Chicago won outright and the teams combined for 51 points. This was two correct selections out of three, not a predicted Bears victory.
The point model projected Seattle 25.93–Chicago 23.87, or Seattle by 2.06 with a 49.80 total. The total was only 1.20 points too low, yet the winner and scoring distribution were badly wrong. Moving from a projected Seattle win by 2.06 to an actual Chicago win by 17 is a 19.06-point margin error. A close total estimate can conceal a large mistake about which team will score the points.
Jacksonville’s sweep and Cincinnati’s scoring miss
Jacksonville’s 42–7 win over Indianapolis delivered JAX to win, JAX −8 and Over 46.5. It joins Green Bay as the week’s second complete sweep. The 35-point margin was much larger than the separate point model’s 6.40-point projection.
The projected score was Jacksonville 28.24–Indianapolis 21.84, totaling 50.08. The actual 49-point total was only 1.08 points lower, but Jacksonville scored 13.76 more than projected and Indianapolis 14.84 fewer. Those errors largely canceled in the total. That is useful evidence against treating total accuracy as a complete measure of score-model quality.
Cincinnati’s 43–36 win over Tennessee produced a different split. The Bengals won, and Tennessee covered +10.5 because Cincinnati’s margin was seven. But the 79-point total overwhelmed the saved Under 46.5 selection by 32.5 points. The point model’s 43.40 combined-point projection was 35.60 too low.
Totals also missed badly in the other direction. Atlanta and Tampa Bay combined for just 16, against an Over 46 selection and a 47.98-point projection. Las Vegas–New York produced 71, defeating Under 45.5. The 5–9 total record includes errors on both high-scoring and low-scoring games, rather than one uniform tendency.
How the individual models compared
Bayesian led winner accuracy at 12–2 in this 14-game sample. Elo, Random Forest and the ensemble each went 10–4.
| Winner model | Correct–incorrect |
|---|---|
| Bayesian | 12–2 |
| Elo | 10–4 |
| Random Forest | 10–4 |
| Weighted ensemble | 10–4 |
| XGBoost | 9–5 |
| Logistic regression | 8–6 |
The ensemble’s weekly log loss improved to 0.58515, from 0.72576 in Week 7, with a Brier score of 0.19859. Lower values indicate smaller probability errors. This was an improvement in probability scoring as well as the winner hit rate; one good week still cannot establish that the percentages are calibrated.
| Market model | Saved spread record | Saved total record |
|---|---|---|
| Logistic / Linear | 7–7 | 6–8 |
| XGBoost | 8–6 | 6–8 |
| Random Forest | 6–8 | 5–9 |
| Ensemble aggregate | 8–6 | 5–9 |
The first row pairs the Logistic spread model with the Linear total model. Elo and Bayesian market proxies are excluded from this trained-model comparison. The ensemble still includes proxy contributions alongside trained spread and total models, and the probabilities remain raw and uncalibrated.
Tuesday’s refreshed monitoring returned to GREEN, with 0.68959 log loss across 42 games from Weeks 6–8. The separate latest-two-week winner window is 18/28, or 64.3%. GREEN means the available measures cleared the existing warning thresholds; it does not establish betting profitability. Settled-bet ROI remains unavailable.
Week 9’s opening look: all 15 games
Here are the early lines we have saved, with Wednesday’s changes alongside them. Tuesday, November 3, at 7:52:42 a.m. CST is our first recorded weekly observation—not a verified sportsbook opener. Wednesday’s snapshot is November 4 at 11:33:55 a.m. CST. Both columns use BetMGM, and neither is a closing line.
Matchups read away–designated home. The spread is the home team’s handicap: a plus sign means that team receives points. Totals are combined points. LAC is the Chargers; LAR is the Rams.
| Matchup | Home spread, Tue → Wed | Total, Tue → Wed |
|---|---|---|
| JAX–BAL | +9 → +10 | 42.5 → 40 |
| CIN–ATL | +3 → +3 | 52 → 52.5 |
| DEN–CAR | +3 → +2.5 | 43.5 → 44 |
| DAL–IND | +6 → +6 | 51.5 → 51.5 |
| NYJ–KC | −9 → −10 | 46 → 47.5 |
| DET–MIA | +7 → +7 | 50.5 → 50.5 |
| CLE–NO | −2.5 → −3 | 43.5 → 42.5 |
| NYG–PHI | +3.5 → +3.5 | 40 → 39.5 |
| LAR–WAS | +6 → +5.5 | 47.5 → 47.5 |
| HOU–LAC | +5 → +5 | 40.5 → 41 |
| LV–SF | −10 → −10 | 51.5 → 52.5 |
| GB–NE | −2.5 → −3 | 47 → 47 |
| ARI–SEA | −11 → −12 | 44.5 → 41.5 |
| TB–CHI | −6.5 → −6.5 | 46.5 → 47.5 |
| BUF–MIN | +3.5 → +3.5 | 46 → 45.5 |
Seven spreads and eleven totals changed. Jacksonville, Kansas City and Seattle each became a one-point larger favorite. New Orleans and New England moved to −3, while Denver and the Rams became half a point smaller favorites. The largest total drops were Arizona–Seattle, down three, and Jacksonville–Baltimore, down 2.5; Jets–Chiefs rose 1.5, the largest increase. Full Wednesday movement report.
Thursday’s Jacksonville–Baltimore matchup is the immediate bridge from this recap: Jacksonville just won 42–7, while Baltimore lost 36–10 at Buffalo. BetMGM moved JAX −9 to −10, cut the total 42.5 to 40, and increased Jacksonville’s moneyline price −475 to −525. That pairs a larger favorite with a lower scoring threshold. It does not prove that last week’s scores—or any particular injury—caused the adjustment.
Injuries, weather and wagering can affect these markets differently. Missing offensive starters can reduce scoring expectations, while absent defenders can increase the opponent’s; stronger wind can affect passing and kicking. Sportsbooks also respond to bets, potential payout exposure and competitors’ prices. We will put the latest team availability and weather beside these moves in the full preview. The saved price snapshots alone cannot identify a cause or justify calling a move “sharp money.”
Sunday night’s Tampa Bay–Chicago line holds at Chicago −6.5, with the total rising 46.5 to 47.5. Monday’s Buffalo–Minnesota spread stays Buffalo −3.5, but its price improves from −102 to +100, while the total falls 46 to 45.5. An unchanged spread can still offer a different payout.
These Wednesday quotes are a market update, not a refreshed recommendation list: the dashboard’s odds and portfolio still reflect Tuesday’s import. New lines do not change the frozen forecasts, and probabilities attached to an earlier handicap or total cannot be transferred to a new threshold.
Tomorrow, Thursday, November 5, we’ll publish the full Week 9 preview, with all 15 picks and a closer look at the week’s primetime games. Today’s record closes Week 8 at ten correct winners, a primetime spread sweep, and nine missed total selections that deserve as much attention as the improved sides.
Sources: the November 4 recap handoff, original Week 8 forecasts, and Tuesday closeout and monitoring record. Point-model errors use full-precision values before rounding. The records grade the original selections and do not assert that any bet was placed.
Written by cresencio
← Back to blog
