2026 NFL Preseason Week 2 score-prediction notebook
Objective and scoring strategy
The contest ranks exact full scorelines first, then exact individual-team scores, then game outcomes, then SSE. I therefore treated late betting totals/spreads as calibrated central estimates but mapped those continuous means onto common discrete football scores (13, 14, 16, 17, 20, 21, 24). Repetition of 20-17 or 17-20 is intentional: most games have totals of 36.5-37.5 and modest spreads, making those pairs plausible scoreline modes. Artificially diversifying scores across games would reduce expected exact hits.
Core market snapshot
A near-cutoff Yahoo board embedded in evidence sr_256041b201e7125908a7aeb9 showed: DEN -6 / 40.5; DET -4.5 / 37.5; CLE -3 / 36.5; IND -3.5 / 36.5; MIN -3.5 / 36.5; NO -3 / 37.5; NYG -3.5 / 36.5; CHI -1.5 / 36.5; PHI -2.5 / 36.5; TB -5.5 / 37.5; ARI -1.5 / 37.5; TEN -4.5 / 37.5. Some other late sources differed by about one point, which is normal in volatile preseason markets.
The Big Lead all-game guide (sr41496503b5fdc01424156910) supplied lineup context and generally leaned to unders: Josh Allen not expected for Buffalo; Atlanta starters sitting; Minnesota's J.J. McCarthy/backup depth receiving work; Tampa Bay starters including Baker Mayfield expected for limited action; Kansas City protecting top talent; and Tennessee favored behind its QB rotation. Covers' all-game win market (sr7277394522433c27fad5fddd) broadly agreed with the selected favorites: Denver 67%, Detroit 64%, Cleveland 56%, Indianapolis 64%, Minnesota 62-64%, New Orleans favored over the Rams, Giants 57%, Bears 53%, Eagles favored, Tampa Bay 67%, Arizona narrowly favored, and Tennessee 64%.
Game-by-game rationale
GB at DEN — DEN 24-17
Late line DEN -6, total 40.5 maps naturally to roughly 17-23/24. Denver has the deeper roster, Bo Nix was expected to play, and Green Bay planned only a short Jordan Love appearance. The 24-17 pair is both market-aligned and composed of highly common football scores.
WAS at DET — DET 20-17
Detroit was favored, with reports that a majority of starters would play while Jayden Daniels' status was uncertain. Tactical Logic AI explicitly projected Detroit 20, Washington 17 (sr_898e5365aed7c6dd963b0802). That projection also fits the market total and common-score lattice, so I retained it despite line variation from DET -1.5 to -4.5.
BUF at CLE — CLE 20-17
Covers' current preview (sr_aec5b9ff2ad9a4e57274a58f) had Cleveland -2.5 and under 36.5, noting Josh Allen was unlikely to play and Cleveland had motivated QB evaluation snaps for Shedeur Sanders and Deshaun Watson. I used 20-17 rather than 20-16 because 17 is a more common exact team total and is only half a point above the listed game total.
ATL at IND — IND 20-13
The late line was Indianapolis -3.5 with total 36.5. Atlanta planned to sit most starters, had a weak backup-QB outlook, and scored only seven in its opener. Indianapolis also had the better QB-depth case. A 20-13 result leans under and allows for Atlanta's offensive floor while preserving the favored outcome.
BAL at MIN — MIN 20-17
Minnesota was around -3 to -3.5 with a 36.5 total and roughly 62-64% win-market probability. J.J. McCarthy and Carson Wentz gave Minnesota the more attractive depth profile against Baltimore backups. The 20-17 score is the strongest common-score representation of a close, low-total home win.
NO at LAR — NO 20-17
This was the hardest side. Multiple expert previews picked the Rams as a value underdog, citing serviceable Rams reserve QBs and a defense that allowed only 12 in the opener. However, the late market still held New Orleans around -3 with a 37.5 total. Since expert underdog picks do not imply the underdog is more likely outright, I followed the calibrated market winner while using the modal 20-17 score pair.
NYG at MIA — NYG 20-16
The Giants were -3.5 with total 36.5, producing implied means almost exactly 20 and 16.5. Miami generated little after its starters' opening touchdown in Week 1, while the Giants had the stronger reserve-QB/depth argument. This is the closest direct mapping from market expectation to discrete football scores.
CHI at CIN — CHI 20-17
The late board moved to Chicago -1.5 and total 36.5. Analysts were split, with some earlier Cincinnati support, but Covers preferred Chicago's Case Keenum-led depth over Cincinnati's Josh Johnson/Sean Clifford rotation. A 20-17 Chicago win reflects both the narrow favorite and under-ish environment.
PHI at NE — PHI 20-17
Current market PHI -2.5 / 36.5. Neither Jalen Hurts nor Drake Maye was expected to play much, if at all. New England had a plausible upset case behind Tommy DeVito, but Philadelphia's Tanner McKee/Andy Dalton pairing and a separate SBG analysis (sr_167c559c8c36b9e3f73b8101) supported Eagles plus under. Selected 20-17.
KC at TB — TB 21-14
Tampa Bay moved sharply to roughly -5.5 after announcing 8-15 snaps for its starters, while Kansas City planned sparse top-unit use. A Covers/Yahoo preview (sr_f052c0bb45867258e026ba4e) favored an under because both teams should run heavily and Florida heat/humidity could slow the second half. The 21-14 result totals 35, gives Tampa a seven-point win, and uses very common scoring combinations.
DAL at ARI — ARI 20-17
The late market was ARI -1.5 / 37.5. Dallas had an upset case through Joe Milton and Sam Howell, and Covers selected Dallas at a 47% market price, but Arizona remained the slightly more likely outright winner. A 20-17 home win is the conservative modal treatment of the near coin flip.
SEA at TEN — TEN 21-17
Tennessee was approximately -4.5 with total 37.5 and around 64% in the win market. Tennessee expected substantial Cam Ward/Will Levis work, while Seattle leaned on Drew Lock and Jalen Milroe. One external projection suggested TEN 24-13, but that margin was much larger than the market. I used 21-17 because both totals are common, the combined 38 is essentially on market, and Tennessee remains the winner.
Assumptions and uncertainty
Preseason results are unusually sensitive to unannounced rotations, joint-practice workload, late scratches, turnovers, and fourth-quarter roster-quality mismatches. Lines from different timestamps varied, so I prioritized the latest broad Yahoo snapshot and used previews as context rather than instructions. No ties were predicted because each individual tie probability is low, although preseason ties remain possible. The final slate primarily follows late market winners and uses common scoring values to maximize exact-team and exact-scoreline probability.
