DeVonta Smith

A.J. Brown was traded to the Patriots in June 2026 — Smith is now the undisputed WR1. He averaged 5–8 TDs and 833–1,196 yards per season as the WR2 next to Brown; these numbers estimate what happens to that rate now that he's the focal target. Brown himself averaged 0.50 TDs and 74.9 yards per game over his last two Eagles seasons — the open question is how much of that vacated production Smith absorbs. Low/Mid/High assume he absorbs 25% / 40% / 55% of it: 10.1 / 11.7 / 13.1 expected TDs and ~1,564 / 1,756 / 1,947 expected yards across the 17-game slate (vs. ~6.2 TDs and 1,004 yards/season historically).

The 40% central estimate stays moderate rather than aggressive because of two other real 2026 facts: Dallas Goedert set a franchise-record 11 receiving TDs in 2025 on his own, independent of Brown leaving, and Saquon Barkley remains the workhorse back under a new, more run-heavy "wide-zone" scheme — both compete for the same red-zone volume and receiving-game touches Smith would otherwise fully inherit.

Left deliberately unresolved: Jalen Hurts' goal-line "tush push" competes for the same 1-yard-line scores (yards, not TDs, are unaffected by this one). His rushing TDs went 10 → 13 → 15 → 14, then dropped to just 8 in 2025 — tracking a tush-push conversion-rate fall from 81% to 63.6%, not a play-calling change. The play stays legal in 2026. Whether that recovers (headwind on Smith's TD odds) or holds (tailwind) is genuinely unclear, so it isn't baked into a direction here.

Independent cross-check, via target share instead of production: the Eagles ran 497 team pass attempts in 2025; Smith drew 22.7% of them, Brown 24.3%, Goedert 15.7%. Redoing the same 25/40/55% absorption in target-share terms (rather than TD-rate or yardage terms) puts Smith at 28.8% / 32.5% / 36.1% target share for 2026 — right in the ~25–32% range real true-WR1s typically run, and above Brown's own 24.3% share even in the low scenario. Two independent methods landing in the same realistic neighborhood doesn't prove the scenarios are right, but it's a real test that could have shown them to be unrealistic, and didn't.

Small print: the low/mid/high scenarios are a reasoned judgment call, not a statistical fit — there's zero historical data of Smith as a true WR1 to fit a model on. TDs are scaled by a ratio (bounded, since probabilities can't exceed 100%); yards are scaled by adding Smith's assumed absorbed yards/game directly, since multiplying an already-elevated yardage estimate by the same ratio produced borderline-record, unrealistic totals — additive is the more stable and semantically honest way to say "he picks up some of Brown's vacated volume." Opponent matchup differences (why Washington's different from Chicago) come from small models trained on his 2021–2025 game log and each 2026 opponent's preseason defensive rating; that part of the estimate is real but weak on its own — the TD model barely beats guessing his flat career rate, and the yards model is similarly modest (R² ≈ 0.04). The WR1 role change is the dominant factor in both sets of numbers, and it's the part that's judgment rather than data.

Two things tried and NOT included, on purpose: opponent red-zone TD% allowed was tested as a real feature (historical data, cross-validated the same way as everything else) and it made both models measurably worse — so it's left out, same as the earlier 7-feature version of the TD model. Man/zone coverage rate and blitz rate (below) are real and current, but only exist as a single 2026 snapshot with no historical seasons behind them to validate against real outcomes — so they're shown as reference only, never folded into a probability.