Thursday Game: BUF vs DET

Pulling a bit of Week 2 context, then a team-by-team read through Vegas and efficiency.

Thursday Night, Highmark. Two 1-0 teams that just survived 36–31 and 31–30. Vegas is pricing a shootout. 

Efficiency is pricing Buffalo as the cleaner Week 1 team and Detroit as the more volatile one. Those two lenses do not fully agree on how the yards show up.

cbssports.com

Buffalo via Vegas

Vegas is not asking Buffalo to be a grind team. It is asking them to win a 30-point game at home.

• -4.5 / 67% / 29.5 points is a clear but not huge favorite. This is not a -10 script.

• 54.5 says both offenses stay on schedule. A 29.5 implied total is an Allen passing game, not 96 rush yards as the engine.

• Your DAP card already agrees with Vegas on BUF shape: Pass 334.6 / Rush 95.7 / R-P 0.90. That is the one side of the slide that matches the board.

• Week 1 already printed the same picture: Allen 334 pass / team 86 rush / 36 points. Forecast was 178/169. Pass was short by 157. Rush was long by 83.

• Lions just played 86 defensive snaps and allowed Shough 410. Vegas is betting Allen gets the same look, only better.
  cbssports.com

Vegas read on BUF: take the points and the passing volume. Do not fade Allen’s yardage because DAP has BUF rush at 96. The market total only works if Buffalo throws.

Detroit via Vegas

Vegas is kinder to Detroit than the raw DAP file was.

• +4.5 / 25 points / 33% is “live dog in a shootout,” not “Gibbs-only, Goff hidden.”

• A 25-point implied total almost never lives at 118 pass yards. That is why the override exists.

• The board moved away from Detroit after Week 1 even though they won. That is the secondary (Joseph/Branch still out) and the 86-snap defensive night, not Gibbs.
  vsin.com

• Total up to 54.5 while the spread lengthens toward BUF means more points, but the market trusts Buffalo to score them.

Vegas read on DET: they score in the low-to-mid 20s. They do not put up 263 rushing yards on the road in a track meet. They have to throw enough to keep Allen off the field.

Team-by-team synthesis

Bills. Vegas and EFF agree. This is Allen’s game. Home, rank-4 efficiency; Lions secondary just got cooked; total is the highest on the board. The card’s BUF 335/96 is the side I would actually use. 

The Week 1 error says if there is another miss, it is more pass, less rush again. Lions. Vegas and EFF agree they are live, and both disagree with the raw 118/263 script. 

Override to 180/175/355 is the version that can sit next to a 25-point implied total without lying. Gibbs can still be the best skill player on the field. He cannot be 75% of the offense on the road against a team that just allowed 334 through the air and will force Detroit to keep up. Game. 

Two lenses, one shape: Buffalo 27–31, Detroit 20–24, ball in the air more than the raw R-P file wanted. Spread is fair at -4.5. 

Total is the fight — 54.5 is rich, but both W1 games already cleared 60 combined points, and the Lions defense is the tired unit on a short week.

Two different machines. Same game. They disagree on who runs.

GSPI is a scoreboard model: spread + implied totals + live QB → a script label → a rush/pass multiplier.

GAME 13 / raw + DET override is a unit model: efficiency, home/away splits, last-3 identity, Week 1 residuals, then a manual Detroit cap.

That is a real test, not two slides of the same idea.

Where they agree

• Buffalo is the favorite. Your card: −4.5, 67% win, proj 29.5–25. GSPI: −4.5, 64/36, implied 29.0–24.5. Same game shape.

• Totals sit in the same band (53.5 vs 54.5). This is a high-scoring opener, not a 41-point grind.

• Gibbs is not a fade. GSPI: three-down, script only a mild drag. Your card: DET RB Above Avg, 175 rush yards after the override.

• Neither system calls this Strong. 4.5 points is not two scores.

Where they fight

Side

GSPI lean

GAME 13 raw + override

Conflict

BUF

Rush ×1.10 / Pass ×0.94

Pass 335, Rush 96, R/P 0.90

Favorite is supposed to run more. Your engine has them throwing.

DET

Rush ×0.92 / Pass ×1.08

Pass 180, Rush 175, R/P 0.97

Dog is supposed to throw more. Override has them running.

GSPI says: if Buffalo leads, Cook’s early downs rise, and Allen’s attempts flatten; Detroit’s path is Goff throwing to keep pace.

GAME 13 says the opposite mix:

• Buffalo’s offense is Easy against a Detroit defense graded Below Avg, and Week 1 already printed P+157 / R−83. The raw engine believes Allen stays in the air even as a home favorite.

• Detroit’s last-3 rush rate is 45.8%; Week 1 was R+45; Gibbs just took 29 carries. The override (Pass 180 / Rush 175) tells the model: don't turn Goff into a deficit machine; keep the Lions’ identity.

That override is the whole experiment. Without it, a chase-script model would spike Detroit pass attempts toward the GSPI ×1.08. You capped them at 180 pass yards so Gibbs does not get scripted off the page.

How to read the disagreement

1. GSPI is conditional on the lead. It only pays if Buffalo is ahead in the second half. If this stays 27–24 into the fourth, both rush leans weaken, and your raw pass-heavy Bills line is the better map.

2. GAME 13 is conditional on identity + matchup grade. BUF Off Easy, DET Off Tough, DET QB/WR/TE all Tough. That is a “Bills throw it, Lions grind it” game even if the spread is only 4.5.

3. Week 1 residuals are doing a lot of work on your card. BUF W1 +157 pass / −83 rush is one game against Houston. DET W1 +45 rush is one game that needed overtime. GSPI is designed to avoid overfitting that sample. Your override is a bet that Detroit’s run rate is structural, not a Week 1 blip.

What would falsify each view on Thursday?

• GSPI looks right if Buffalo’s rush share climbs from Week 1’s 40% toward the mid-40s or 50s, and Detroit’s pass share rises as they trail.

• GAME 13 looks right if Allen still clears ~30 dropbacks and ~300 pass yards while winning, and Gibbs still clears ~25 carries / 150+ rush even as a road dog.

• Both miss if it becomes a 38–35 track meet with no one controlling the clock (total hits, mix is noise).

• Both miss if Buffalo sits on a two-score lead and Detroit abandons the run — then GSPI’s DET pass lean finally shows up and your 175-rush override is the miss.

Practical split for the test

Use GSPI for direction of the lead (Cook vs. Goff attempts if BUF is up).

Use GAME 13 for who you actually roster: Allen passing remains live; Gibbs rushing remains live; do not fade Detroit’s run game just because they are +4.5.

The interesting number on your graphic is not the 67% win rate. It is DET OVERRIDE PASS 180 RUSH 175. That is you choosing identity over script.