RESEARCH NOTEBOOK — NFL Preseason Week 2 (Aug 22-24, 2026) final score predictions.
PRIMARY EVIDENCE BASES
- Covers/Yahoo "NFL AI Predictions for Preseason Week 2" (ChatGPT-trained model with playing-time + matchup data). Full score table recovered for all 12 slate games:
DEN 24-13 (GB-DEN); WAS 17-16 vs DET; BUF 23-16 vs CLE; IND 20-13 (ATL-IND); BAL 20-17 vs MIN; LAR 24-20 (NO-LAR); NYG 17-13 vs MIA; CHI 23-17 vs CIN; NE 20-13 (PHI-NE); TB 23-16 (KC-TB); DAL 23-20 vs ARI; TEN 20-16 (SEA-TEN).
- Polymarket market probabilities: DEN 70%, DET 66%, MIN 63%, CLE 59%, IND 64%, TB 69%, ARI 54%, TEN 66%.
- Market lines (Fox/DK/Big Lead/thelines/WindyCity): DEN -4.5(-5.5)/40.5; DET -4.5/37.5; CLE -1.5/36.5; IND -3.5/-4.5 (40.5 Big Lead, 36.5 Kalshi); MIN -2.5/-3.5/36.5; NO -1.5(-3)/37.5; NYG -1.5/36.5; CHI -1.5/-2.5/36.5; PHI -1.5/36.5; TB -5.5/37.5; ARI -2.5/38.5; TEN -4.5/38.5.
- Covers moneyline win-probabilities: Broncos 67%, Lions 64%, Browns 56%, Colts 64%, Vikings 64%, Rams 41%, Giants 57%, Bears 53%, Patriots 45%, Bucs 67%, Cowboys 47%, Titans 64%.
- QB rotation reporting: DEN Bo Nix debut + Stidham/Ehlinger (GB Love 1 series then McCord/Tyrod); DET Goff + starters expected (WAS sits Daniels, Hartman/Kaliakmanis); CLE Shedeur Sanders starts (BUF Allen out, Kyle Allen/Buechele); IND Leonard 1H then Richardson 2H (ATL starters sit); MIN Murray/McCarthy/Wentz deep room (BAL resting, Fagnano); LAR sits Stafford, Simpson/Bennett (NO Shough/Rattler/Wilson, Jordyn Tyson out); NYG Haener/Dart (MIA Cam Miller/Gronowski); CHI Keenum (Bagent hamstring, CIN Josh Johnson/Clifford); TB Baker Mayfield debut (KC Fields/Nussmeier); TEN Ward/Levis (SEA Lock/Milroe); ARI rookie Love out (DAL rests starters).
DECISION FRAMEWORK
- Winner picks anchored to sharp-market (Polymarket) + consensus: DEN, DET, CLE, IND, MIN, TB, ARI, TEN.
- Four remaining: NYG (line + consensus), CHI (line + consensus), LAR (AI/Big Lead/Covers over market NO), NE (AI/experts over market PHI). LAR and NE are recognized coin-flips where I leaned to the AI/qualitative side.
- Where my winner matched the AI model, I adopted its exact score (DEN 24-13, IND 20-13, LAR 24-20, NYG 17-13, NE 20-13, TB 23-16, TEN 20-16), adjusting CHI 23-17 -> 20-17 given the joint-practice/under lean and 36.5 market total.
- For the four market-vs-AI disagreements (DET, CLE, MIN, ARI) I picked market winners with LOW totals consistent with joint-practice under leans, NOT the AI's underdog calls.
KEY UNCERTAINTIES / ASSUMPTIONS
- Preseason exact scores are extremely noisy; the AI model is one well-sourced reference, not ground truth.
- Multiple sources cite 10 joint practices -> "tight, under-ish" games (BAL-MIN, BUF-CLE, NO-LAR, NYG-MIA, CHI-CIN, PHI-NE, SEA-TEN), supporting modest totals.
- Assumed no last-minute injury/rotation changes materially flipping QB plans after the Aug 21 evidence snapshot.
- NO@LAR and PHI@NE are near coin-flips; going with Rams and Patriots.
FINAL SCORES GB 13 @ DEN 24; WAS 16 @ DET 20; BUF 16 @ CLE 20; ATL 13 @ IND 20; BAL 16 @ MIN 20; NO 20 @ LAR 24; NYG 17 @ MIA 13; CHI 20 @ CIN 17; PHI 13 @ NE 20; KC 16 @ TB 23; DAL 17 @ ARI 20; SEA 16 @ TEN 20.
