San Francisco 49ers at Los Angeles Rams. Pick: San Francisco 49ers +3.5. Lean: Under 48.5.

Spread.Take San Francisco +3.5, available at DraftKings at -102. The model's Rams -5.2 includes 2.122 points of home advantage for a game actually played in Melbourne. Removing that allowance reduces the raw Rams margin from 4.6 to approximately 2.5; that is a diagnostic, not a recalibrated forecast. Getting the hook above three therefore matters more than the original 1.7-point apparent Rams edge. Donald's confirmed absence helps San Francisco, although McVay expects Garrett to play, and Kittle's limited participation prevents assuming full offensive strength. The opposing case is substantial: Los Angeles had markedly better 2025 efficiency in the dossier, and its pass rush remains dangerous. The market already knows the venue and Donald news; neither is a discovered market oversight. DraftKings' larger SF handle share than ticket share is supporting context, not proof of sharp action. My lean rests on the price and neutral venue, with incomplete final injury designations keeping confidence modest.

Total.Take under 48.5. Applying the dossier's specified 1.2-point bias correction puts the model at 47.75, a modest 0.75-point cushion before correcting its erroneous dome input. I would not assign an exact additional deduction without rerunning the model. The over case is credible: returning starting quarterbacks, productive offenses, Donald's absence and mild forecast conditions support efficient scoring. Defensive installation changes could also create explosives. Against that, Kittle's unresolved workload adds offensive uncertainty, while FanDuel reports a decline from 49.5 to 48.5 and DraftKings shows substantial under handle. Those moves mean some under value is already gone. I verified no current pace advantage and impose no automatic Thursday penalty on a season opener. This is a small price-based under lean, not a weather-driven position.

Where our number looks wrong.The Rams home-field allowance is inappropriate for Melbourne, and the dome designation is wrong. Removing home advantage makes the raw spread approximately Rams -2.5 before recalibration, erasing the advertised Rams value. The totals model's stated 1.2-point upward bias reduces 48.95 to 47.75 before correcting the venue. Injury blindness and a changed SF defensive staff further undermine precision; neither model has demonstrated an edge over market prices.

Spread.Our number (Rams -5.2) is unusable as written: it bakes in 2.122 points of Rams home field for a game at the Melbourne Cricket Ground, a neutral site in Australia (NFL.com international page, 9/9). Strip the home field and our own raw team-quality gap is 2.48 (LAR 3.45 vs SF 0.97), i.e. a neutral number near Rams -2.5/-3 - under the market, not over it. The model's "Rams undervalued" read inverts once the venue is correct. The market moved from -2.5/-3 to -3.5 on three inputs, per BetMGM's senior trading manager (SI, 9/2): San Francisco's WR/TE injuries, the Myles Garrett trade, and Aaron Donald's return. Two have decayed since. Donald is OUT and did not travel (therams.com, 9/8). Garrett has barely practiced in a month with knee swelling and was a non-participant Tuesday and Wednesday (49ers Webzone 9/8; DK Network 9/9), and Los Angeles sent Jared Verse to Cleveland to acquire him, so the edge room is a swap, not a pure add. San Francisco's receiving corps also isn't bare: Mike Evans is healthy and set to debut. Taking the points through 3.

Total.Our model says 48.95, but it runs about 1.2 points hot on holdout; debiased it sits near 47.8 - below the market. Our own number leans under once its documented bias is applied. It also applied a dome layer to an open-air cricket ground, though conditions are benign (clear, 17C max, near windless, dry surface - Weatherzone, 9/9), so the venue is neutral for scoring rather than suppressive. The lean is personnel and body clock, not weather: a 10:35am local kickoff is unprecedented for both teams, it is Week 1 for two timing-dependent offenses, and San Francisco's pass-catching group is thin - Pearsall on season IR, Aiyuk off the roster, Kirk out to at least Week 5, and Kittle on a limited workload in his first game since a January Achilles tear.

Where our number looks wrong.Three concrete errors, one of them decisive. (1) Venue: the dossier has roof=dome and applies home field 2.122 to the Rams, but this game is at the Melbourne Cricket Ground - an open-air neutral site 8,000 miles from Los Angeles. Any 'gap' reading built on -5.2 is measuring a venue the game isn't being played in. (2) The frozen 2026-08-19 preseason fit cannot see the two largest roster events on the board: the Rams acquiring Myles Garrett (for Jared Verse) and Aaron Donald unretiring, nor San Francisco replacing Aiyuk/Pearsall with Mike Evans. The rating gap of 2.48 is a 2025 artifact. (3) The empty injuries table means the model prices Donald in, Donald out, Garrett rusty and Kittle at 60% identically. Separately, San Francisco's 12-5 on a +3.9 point differential is close-game overperformance that should regress - that argues for the Rams and is the honest counterweight to my side; it is roughly the one thing the rating gap does capture.

Spread.Our raw margin (Rams -4.6 pre-calibration, -5.2 after) already embeds a 2.122-point home-field constant on top of a 2.48-point rating gap. This is not a true home game: it's a neutral-site opener at Melbourne Cricket Ground, with the Rams the designated home team only for NFL scheduling (49ers.com, 2026-02-05; NFL.com Melbourne game page). Strip that constant and re-run the same calibration stretch and our fair number lands near Rams -2.8, effectively matching or bettering the market's -3.5 in San Francisco's favor. Injury news roughly offsets: Rams DT Aaron Donald is ruled out and did not travel to Australia (49erswebzone, 9/8), hurting L.A.'s interior rush on Purdy, while SF DT Alfred Collins tore his patellar tendon in Melbourne and is done for the season (Heavy.com, 9/9), thinning SF's front against the Rams' run game. Sharp moneyline action has still pushed the number to -3.5 from -2.5/-3 (VegasInsider, 9/9), so I'm not fading the market hard, just buying the extra half-touchdown on a 'home' team that has no real home edge here.

Total.Our totals model's own holdout shows it runs about 1.2 points above the closing number, so its 48.95 call likely overstates true expectation by a similar margin, closer to the market's 48-48.5. The model also assumes a domed, artificial-turf setting (layers.dome=1) for what is actually an open-air grass game at the MCG; the forecast is clear, ~17C/63F, near-windless (TheUSSports.com, 9/9), mild, but not the fast, contained environment the model priced in. Aaron Donald's absence should ease interior pressure on Purdy, but Alfred Collins' season-ending injury thins SF's run defense, and both offenses face short-week international travel and first-game rust in a division matchup that has historically trended lower-scoring. The market total has already drifted from ~49 at open down to 48-48.5 on this news. I'll take the under at 48.5.

Where our number looks wrong.The 2.122-point home-field constant assumes a genuine home crowd/travel edge that doesn't exist at a neutral-site game in Melbourne; baking it in overstates the Rams' true number by roughly 2 points once it runs through calibration. The dossier also tags roof as 'dome' and surface as 'matrixturf', but this is an outdoor grass game at the MCG, the totals model's dome layer input doesn't reflect the real venue. And per the model's own disclosure, the injuries table holds zero NFL rows, so it's blind to Donald's absence, Collins' season-ending injury, and the Bosa/Kittle workload management, all of which are actually moving this line.