Washington Commanders vs Philadelphia Eagles Prediction & Betting Pick

=== NFL EDGE ENGINE v3.1 — WASHINGTON COMMANDERS @ PHILADELPHIA EAGLES === NFL Week 1 · Sunday September 13, 2026 · 4:25 PM ET / 1:25 PM PT · Lincoln Financial Field, Philadelphia PA · FOX Run timestamp: 2026-09-13 04:43 PDT / 07:43 ET · Time to kickoff: 8 h 42 m · Activation mode: RUN FULL MODEL Source files read before start: v3.1 spec (502 lines) · run-discipline v2 (308) · calibration v1 (192) · marketmath.py · clampdiagnostic.py — all five verified on disk, both scripts re-copied from the read-only mount (checksums match).

C-7 · PRIOR-RUN GRADING

No newly resolved runs. The two 1:00 PM ET games run earlier this session (ATL @ PIT, BAL @ IND) have not kicked off. NE @ CLE grading from the prior session stands: Browns 37, Patriots 13, no capital at risk, all three declined sub-1% leans won, V-4 tracking 6-for-6.

Clock note (N-9): the device reads 04:43 PDT here against 04:26 PDT on the prior turn — forward by 17 minutes this time, having gone backwards between the two turns before that. Recorded as observed; not differenced.

PRE-FLIGHT (14.1)

Phase REGULAR ▢✅ · Home PHI ▢✅ · Away WAS ▢✅ · Date 2026-09-13 ▢✅ · Date guard ▢✅ · Time to kickoff recorded ▢✅ · Blocks inventoried ▢✅ · Primary game page resolved and fetched FIRST ▢✅ (FOX boxscore-11049, opened before any market was graded)

FOX UTC artifact — fifth confirmed instance. Header rendered "TODAY, 8:25PM"; EVENT INFO read Sep 13. Canonical URL, , meta-dc.title and og:url all carry sep-13-2026 and GOVERN. 20:25 UTC = 4:25 PM ET. No wrong-slate discard taken on the header.

[0] DATA AUDIT GATE 0 — PRIMARY SOURCE CONFIRMATION

=== NFL EDGE ENGINE v3.1 — WASHINGTON COMMANDERS @ PHILADELPHIA EAGLES === NFL Week 1 · Sunday September 13, 2026 · 4:25 PM ET / 1:25 PM PT · Lincoln Financial Field, Philadelphia PA · FOX Run timestamp: 2026-09-13 04:43 PDT / 07:43 ET · Time to kickoff: 8 h 42 m · Activation mode: RUN FULL MODEL Source files read before start: v3.1 spec (502 lines) · run-discipline v2 (308) · calibration v1 (192) · market_math.py · clamp_diagnostic.py — all five verified on disk, both scripts re-copied from the read-only mount (checksums match).

---

# C-7 · PRIOR-RUN GRADING

No newly resolved runs. The two 1:00 PM ET games run earlier this session (ATL @ PIT, BAL @ IND) have not kicked off. NE @ CLE grading from the prior session stands: Browns 37, Patriots 13, no capital at risk, all three declined sub-1% leans won, V-4 tracking 6-for-6.

*Clock note (N-9): the device reads 04:43 PDT here against 04:26 PDT on the prior turn — forward by 17 minutes this time, having gone backwards between the two turns before that. Recorded as observed; not differenced.*

---

# PRE-FLIGHT (14.1)

Phase REGULAR ▢✅ · Home PHI ▢✅ · Away WAS ▢✅ · Date 2026-09-13 ▢✅ · Date guard ▢✅ · Time to kickoff recorded ▢✅ · Blocks inventoried ▢✅ · Primary game page resolved and fetched FIRST ▢✅ (FOX boxscore-11049, opened before any market was graded)

FOX UTC artifact — fifth confirmed instance. Header rendered "TODAY, 8:25PM"; EVENT INFO read Sep 13. Canonical URL, `<title>`, `meta-dc.title` and `og:url` all carry sep-13-2026 and GOVERN. 20:25 UTC = 4:25 PM ET. No wrong-slate discard taken on the header.

---

[0] DATA AUDIT

GATE 0 — PRIMARY SOURCE CONFIRMATION

| Block | Primary per spec | Fetched? | Actual source used | Time | Sub? | |---|---|---|---|---|---| | 0 date guard | phase/date fix | Y | FOX boxscore-11049 canonical + meta; corroborated philadelphiaeagles.com, ESPN, NFL.com, DirecTV | 04:5x PDT | N | | A game overview | FOX game page | Y | FOX boxscore-11049 | 04:5x PDT | N | | B odds current | user-pasted DK | N | DK-attributed at two outlets (FOX "How to Watch" today; SI today) + FOX game-page module + FanDuel + ESPN + thespread | today / 1 d | Y — no user paste; spec secondary satisfied, but the two DK-attributed outlets disagree with each other (see discards) | | C odds open/movement | VSiN line history | N | SBD movement tracker 9/12; FOX "How to Watch" today; thespread today; SI today; cleatz 9/8 | 9/8–9/13 | Y — VSiN not attempted; four same-day sources produce mutually contradictory movement, see below | | D public splits | DK Network splits | N | cleatz 9/8 full bets+handle board; cleatz 9/5 ML handle | 9/5–9/8 | Y — DKN widget server-renders MLB only (fifth confirmation); no post-injury-report split located → Block D is stale by 5 days and flagged | | E weather | search "[stadium] weather [date] kickoff hour" | Y | FOX Weather module (hourly) + ESPN header | today | N — outdoor venue, thresholds evaluated | | F rosters (truth file) | FOX `-team-roster` both teams | N | Commanders official Week-1 depth chart (offense, full) + both official injury reports + Eagles beat naming | 9/9–9/13 | Y — run-budget; WAS side complete, PHI skill set named from reports/props | | G depth chart / starters | Rotowire lineups + coach quotes | N | Commanders official Week-1 depth chart + Quinn presser 9/11 + Sirianni/Fangio pressers | 9/9–9/12 | Y — Rotowire discard-no-retry standing rule; user directive: published/projected lineup counts as PASS | | H injuries both teams | official league injury report | Y | commanders.com three-day participation grid (team official) + philadelphiaeagles.com Friday report; corroborated ESPN, NBC Sports Philly, Inquirer, Bleeding Green Nation, PFT, The Athletic via Yahoo | 9/11 final | N | | I team + player stats | FOX team-stats | Y | FOX 2025 all-32 defensive tables + PhillyVoice 2025 leaders + Yahoo 10-stats + commanders.com preview | 2025 final | N | | J head-to-head | game-page H2H module | Y | FOX boxscore-11049 H2H module (five meetings) | 04:5x PDT | N | | K recent games | game-page RECENT GAMES module | Y | FOX boxscore-11049 (3 rows/team) + H2H module for tier-3 | 04:5x PDT | N | | L props board | user-pasted DK | N | NBC Sports Philadelphia full board 9/11 + BetMGM full ATD/rush/rec board 9/9 + phlsportsnow today + SBD 9/7 | 9/7–9/13 | Y — no user paste; four book-attributed boards, each staleness-tagged |

Circuit breaker: 6 blocks show `Fetched? = N` with `Sub? = Y` — TRIPPED and walked for the third consecutive run. Walk result: B, C, F, G and L were all satisfied at spec-secondary level or better; D is the one genuine loss — no post-Friday split exists in any source opened, so the splits are five days stale and predate both final injury reports. Grading proceeds with Block D explicitly downgraded. See N-5 — a breaker that trips every run is not discriminating.

DISCARDS AND CONFLICTS

1. ⚠️ BLOCK C IS INTERNALLY CONTRADICTORY — four same-day sources, four different movement stories.

| Source | Age | Reported open | Reported current | Direction | |---|---|---|---|---| | FOX Sports "How to Watch" (DraftKings) | today | PHI −5.5 | PHI −4.5 | toward Washington | | SportsBettingDime odds tracker | 1 d | PHI −4.5 | PHI −5.5 | toward Philadelphia | | thespread.com | today | PHI −4 | PHI −5 | toward Philadelphia | | SI (DraftKings) | today | PHI −5.5 | PHI −5.5 | "neither the spread nor the total has moved" |

FOX and SBD report the exact same two numbers in the exact opposite order. Two of these sources (FOX, SI) both claim to be quoting DraftKings on the same day and give different numbers for it. The movement read is not establishable from the available sources and is coded accordingly. No RLM determination is made in either direction — see §6. New ledger item.

2. Reported opening totals also conflict: 47.5 (SBD, thespread), 46.5 (FOX today), 48.5 (ScoresAndOdds 8/27, oddsshopper-class early captures). What every source does agree on: the total has fallen roughly 2.5–3.5 points to 44–44.5. That much is robust and is used; the specific opener is coded SF.

3. FOX game page's own INJURIES module vs its own PLAYER NEWS feed — F-48, fifth instance. The module lists TE Eli Stowers as Questionable; the player-news feed on the same page reports "The Eagles placed Stowers (quadriceps/hamstring) on injured reserve Saturday." IR means unavailable, not questionable. The Eagles also elevated WR Britain Covey from the practice squad Saturday, which corroborates a Saturday roster move. The player-news feed is treated as the fresher truth: Stowers OUT (IR), superseding Friday's Questionable. Immaterial to the game (fourth tight end) but the module contradiction is logged.

4. cleatz splits board (9/8) and ML handle (9/5) — retained but explicitly stale: both predate Friday's injury reports and the late total move. Used for direction of exposure only, never as a current read.

5. CBS "NFL betting splits" article referencing "Cowboys at Eagles" as the season's first game — inconsistent with the 2026 schedule (SEA/NE opened 9/9). Prior-season article. Discarded whole.

RCF STATUS — FALSE

No condition met. Starters published on both sides; no preseason rotation; Washington's offensive line is fully named (Coleman–Paul–Allegretti–?–?) with Allegretti a full participant all week; no QB1 contradiction.

V-1 / F-38 compliance: RCF FALSE, cap not in play. Had it been TRUE it would not have been applied to Philadelphia's probability (65.8% uncapped). No cap-manufactured edge exists in this run.

DEGRADED CONSEQUENCES APPLIED

| Missing / degraded | Consequence applied | |---|---| | C (contradictory) | RLM = NOT DETERMINABLE; movement read reduced to the one robust claim (total fell ~3 pts); opener coded SF | | D (5 days stale) | Book Exposure read carries an explicit staleness caveat; splits may not corroborate any DOG VALUE trigger | | F (PHI side) | Eagles skill pool named from reports and prop boards; tier-1 rates missing for Goedert, Wicks → `(la)` | | B (no user paste) | Satisfied at spec-secondary level → no Lean cap; but the DK-vs-DK conflict is logged | | L (no user paste) | Four boards adopted, each staleness-tagged |

---

[1] GAME HEADER

• League/Event: NFL Week 1, 2026 regular season · NFC East divisional opener
• Matchup: Washington Commanders (0-0) @ Philadelphia Eagles (0-0)
• Kickoff: Sun Sep 13, 2026 · 4:25 PM ET / 1:25 PM PDT
• Venue: Lincoln Financial Field, Philadelphia — outdoor
• Broadcast: FOX
• Coaching: Nick Sirianni (PHI, HC) with new OC Sean Mannion and DC Vic Fangio · Dan Quinn (WAS, HC) with new OC David Blough and new DC Daronte Jones
• Weather: ~82–83 °F, wind 3 mph ESE, no precipitation signal (see §7)

---

[2] QB COMPOSITE (descriptive, non-predictive)

| QB | Team | 2025 season-labeled line | |---|---|---| | Jalen Hurts (starter) | PHI | 64.8% comp, 3,224 pass yds, 7.1 Y/A, 25 pass TD (career high), 6 INT, 32 sacks, 98.5 rating. FOX FACTS: failed to exceed 7.5 rushing attempts in 11 of his last 12 games — 5.6 rush att/game | | Jayden Daniels (starter) | WAS | 60.6% comp, 1,262 pass yds, 6.7 Y/A, 8 TD, 3 INT, 18 sacks — in SEVEN injury-plagued appearances. 2024 Offensive Rookie of the Year; shut down for the rest of 2025. Missed both Eagles games last season | | Marcus Mariota (QB2) | WAS | Full participant all week (knee), available | | Athan Kaliakmanis (QB3) | WAS | depth chart |

Non-predictive display only.

⚠️ Gate 3 note that matters downstream: Daniels' 2025 sample is 188 pass attempts — below Part 10.2's 200-attempt gate. His tier-1 rate is blocked by the spec, by twelve attempts. See §18.

---

[3] TEAM METRICS SNAPSHOT — 2025 regular season (non-predictive)

| Metric (2025 REG) | WAS | rank | PHI | rank | |---|---|---|---|---| | Record | 5-12 | — | 11-6, lost Wild Card to SF | — | | Points/g | 20.9 | — | 22.3 | — | | Points allowed/g | 26.5 | 27 | 19.1 | 5 | | Total yds allowed/g | 384.3 | 32 — LAST | 314.2 | 13 | | Pass yds allowed/g | 242.5 | 28 | 189.8 | 8 | | Opp completion % allowed | — | — | 56.8% | — | | Rush yds allowed/g | 141.8 | 30 | 124.4 | 22 | | Rush yds allowed, YPC | 4.8 (2,411 on 504) | — | 4.3 (2,115 on 494) | — | | Red-zone TD % (offense) | — | — | 70.5% | 1st in NFL | | Red-zone TD % allowed | 67.7% | 31 | 53.1% | 9 | | Rushing offense | 134.7 yds/g | 4th | 116.9 yds/g | — | | Passing offense | 184.1 yds/g | — | — | — | | Turnover differential | −13 | 31 | Hurts 25 TD / 6 INT | — | | Time of possession | 28:26 | 28 | — | — |

This is the widest quality gap of the four games run this session. Philadelphia was 5th in scoring defense and 1st in red-zone touchdown rate; Washington finished dead last in total defense and 31st in red-zone defense. Washington's one strength — 4th-ranked rushing offense — runs into a Philadelphia front that was 22nd against the run, which is the one seam.

Personnel turnover, both sides, which is what the total drop is about: - PHI: A.J. Brown gone (to New England) — DeVonta Smith is now the unquestioned WR1. Brandon Graham retired. Jonathan Greenard signed (and is OUT, §5). The 2025 ground game collapsed from 3,048 yards on 621 carries (2024) to 1,988 on 478, with Barkley's YPC falling 5.8 → 4.1. New OC Sean Mannion after Kevin Patullo's play-calling was widely blamed for 2025. - WAS: Stefon Diggs, TE Chig Okonkwo, RB Rachaad White and rookie LB Sonny Styles added; Deebo Samuel returned to San Francisco; Odafe Oweh starts opposite K'Lavon Chaisson; three new faces in the secondary; both coordinators replaced.

---

[4] PREVIOUS 5 GAMES / TEAM — phase-general backfill ladder

Tier 1 (current season, current phase): N/A (SI — Week 1, both clubs 0-0 in regular-season play). Tier-2 table follows immediately.

WASHINGTON

| # | Tier | Date | Opp | Result | ATS | O/U | |---|---|---|---|---|---|---| | 1 | T2 · 2026 PRESEASON | 8/28 | @ BAL | L 3-41 | +4.5 | O 34.5 | | 2 | T2 · 2026 PRESEASON | 8/22 | @ DET | L 13-17 | +1.5 | U 36.5 | | 3 | T2 · 2026 PRESEASON | 8/14 | vs MIA | W 20-7 | +3.5 | U 36.5 | | 4 | T3 · 2025 REG Wk18 | 1/4/26 | vs PHI | W 24-17 — *against Philadelphia's backups* | SF | SF | | 5 | T3 · 2025 REG Wk16 | 12/20/25 | vs PHI | L 18-29 | SF | SF |

PHILADELPHIA

| # | Tier | Date | Opp | Result | ATS | O/U | |---|---|---|---|---|---|---| | 1 | T2 · 2026 PRESEASON | 8/28 | vs CIN | L 13-30 | −3.5 | O 36.5 | | 2 | T2 · 2026 PRESEASON | 8/22 | @ NE | L 21-24 | −1.5 | O 37.5 | | 3 | T2 · 2026 PRESEASON | 8/15 | @ BAL | L 7-24 | −3.5 | U 37.5 | | 4 | T3 · 2025 POSTSEASON | — | vs SF (Wild Card) | L — score SF | SF | SF | | 5 | T3 · 2025 REG Wk18 | 1/4/26 | @ WAS | L 17-24 (backups) | SF | SF |

Philadelphia went 0-3 in the 2026 preseason and was outscored 41-84. Washington went 1-2 and was outscored 36-65. Both preseason bodies of work are ledger-banned as inputs (Part 6) and printed here as season- and phase-tagged context only, per C13.

HEAD-TO-HEAD (Block J), five meetings, all printed because the sample is unusually rich:

| Date | Result | Note | |---|---|---| | 1/4/26, Wk 18 | WAS 24, PHI 17 | Philadelphia rested starters; Washington was 5-12 | | 12/20/25, Wk 16 | PHI 29, WAS 18 | — | | 1/26/25, NFC Championship | PHI 55, WAS 23 | — | | 12/22/24, Wk 16 | WAS 36, PHI 33 | — | | 11/15/24, Wk 11 | PHI 26, WAS 18 | — |

Roster turnover since each meeting is heavy on both sides (A.J. Brown, Brandon Graham, Deebo Samuel all gone; Diggs, Okonkwo, White, Oweh, Greenard, Styles all new). Context only. Not a ledger input.

---

[5] NEWS / INJURIES — official reports govern (Block H, REGULAR hierarchy)

WASHINGTON — final report Friday 9/11. Dan Quinn: "everyone on the 53-man roster is cleared to play." Full three-day participation grid, from the team's own site:

| Player | Injury | Wed | Thu | Fri | Final | |---|---|---|---|---|---| | OLB K'Lavon Chaisson | Knee | Limited | Limited | Limited | cleared, no designation | | RB Jacory "Bill" Croskey-Merritt | Groin | Limited | Limited | Limited | cleared | | OLB Odafe Oweh | Calf | Limited | Limited | Limited | cleared | | DT Daron Payne | Knee | — | Limited | Limited | cleared | | C/G Nick Allegretti | Calf | Full | Full | Full | available | | DT Javon Kinlaw | Back | — | Full | Full | available | | LB Jordan Magee | Chest | Full | Full | Full | available | | QB Marcus Mariota | Knee | Full | Full | Full | available | | DE Charles Omenihu | Wrist | Full | Full | Full | available | | TE Ben Sinnott | Oblique | Full | Full | Full | available |

Zero OUT, zero QUESTIONABLE. Chaisson and Oweh together produced more pressures than any single player on Washington's 2025 roster — both are available. The FOX module's "No players are injured" is correct for Washington this time.

PHILADELPHIA — final report Friday 9/11: - OUT: EDGE Jonathan Greenard (pectoral) — the club's headline free-agent defensive addition. Missed all of training camp after an offseason weight-training pec strain (no surgery); limited Wed/Thu/Fri; said himself he is "not yet 100%." Fangio: they have to be "careful" with him. - QUESTIONABLE Friday → superseded: TE Eli Stowers (hamstring) — placed on injured reserve Saturday per the FOX player-news feed; WR Britain Covey elevated from the practice squad the same day. Fourth tight end; immaterial. - Rest of the 53-man roster full-go. - Context: Brandon Graham retired. The Week-1 edge rotation is Nolan Smith and Jalyx Hunt starting, with A.J. Epenesa and Arnold Ebiketie behind. Jalen Carter and Jordan Davis anchor the interior.

Net read: Washington is 100% healthy for the first time in a year; Philadelphia is missing its marquee edge addition but starts its intended pair anyway. Both were settled Friday, before any of the market activity that can be dated.

---

[6] PUBLIC & ODDS BOARD

| Team | Spread (live band) | ML | %Bets (spread) | %Handle (spread) | Implied team total | Vig-free mkt % | Model win% | |---|---|---|---|---|---|---|---| | WAS (away, dog) | +4.5 / +5 / +5.5 / +6 | +180 to +205 | 31% *(5 d stale)* | 8% *(5 d stale)* | 19.50 | 32.43% | 34.18% | | PHI (home, fav) | −4.5 / −5 / −5.5 / −6 | −218 to −250 | 69% | 92% | 25.00 | 67.57% | 65.82% | | Total | 44 / 44.5 | O −112 / U −108 | 45% Over | 15% Over | — | O 50.43% / U 49.57% | O 50.19% / U 49.81% |

GATE 5 — MARKET AUDIT TRAIL

Every quote, source, timestamp, disposition:

| Market | Quote | Source | Age | Adopted? | Reason if discarded | |---|---|---|---|---|---| | Spread | PHI −5.5 (open), total o46.5 | ScoresAndOdds board | 17 d | No | Freshness — opener reference only | | Spread | PHI −4.5, 69% bets / 92% handle | cleatz | 5 d | No (as price) | Freshness; retained for exposure direction | | Spread | PHI −4.5 | BetMGM via Bucks County | 2 d | Borderline | superseded by same-book same-day readings | | Spread | PHI −5.5, ML −240/+195, O/U 44.5 | FOX Wk1 odds article | 2 d | Borderline | superseded | | Spread | WAS +5.5 (−115) / PHI −5.5 (−105), ML +194/−235, total 44.5 (−112/−108) | FanDuel via SI | 4 d | Yes (prices) | — | | Spread | PHI −4.5, total 44.5, ML −228/+188 | DraftKings via FOX "How to Watch" | today | Yes | — | | Spread | WAS +5.5 (−112) / PHI −5.5 (−108), ML +180/−218 | DraftKings via SI | today | Yes | ⚠️ conflicts with the line above on the same book, same day | | Spread | WAS +5.5 (+204) / PHI −5.5 (−249), total 44.5 | FOX game-page module | today | Yes | — | | Spread | PHI −5 (−115) / WAS +5 (−105), total 44, ML −240 | thespread.com | today | Yes | — | | Spread | PHI −6, ML −250/+205, total 44.5 | ESPN betting board | 1 d | Yes | — | | ML | PHI −218, 89% of handle | cleatz | 10 d | No | Freshness — but see the staleness note below |

ML OUTLIER SCREEN — `market_math.py`, fresh quotes only, deduped:

``` dog +188 / fav -228 -> dog 33.312% hold 4.23% dog +204 / fav -249 -> dog 31.556% hold 4.24% dog +205 / fav -250 -> dog 31.461% hold 4.22% dog +180 / fav -218 -> dog 34.252% hold 4.27%

median dog probability : 32.434% range : 31.461% to 34.252% (spread 2.792 pts) no quote exceeds the ~2-point outlier threshold ```

This is the widest moneyline dispersion measured in four runs — 2.79 points — and the largest single deviation from the median is +1.82, just inside the ratified 2.0 threshold. The quote responsible is +180 / −218 (SI, today), which is numerically identical to cleatz's ten-day-old −218 capture. That is the signature of a republished stale price rather than a genuinely soft book.

⚠️ Freshness note that cuts the other way from 2026-09-09. Excluding the suspected-stale quote moves the vig-free dog probability from 32.434% to 31.556% — a 0.88-point shift that makes the dog moneyline edge larger, not smaller (+1.75% → +2.62%). Staleness is not only an edge-manufacturing risk; here it is suppressing an apparent edge. The quote cannot be proven stale, so the anchor is reported both ways and the market is declined under both. F-45 should be reworded to cover both directions.

Consensus anchor, stated as numbers: S = 5.5 (PHI, median of the four live numbers), T = 44.5, vig-free ML PHI 67.57% / WAS 32.43%.

⚠️ FOUR LIVE SPREAD NUMBERS — the widest posted disagreement of the session — and the sanity check rejects half of them:

``` Phi(4.5/13.5) = 0.6306 vs fair ML 0.6757 -> gap -4.51 pts FAIL Phi(5.0/13.5) = 0.6444 vs fair ML 0.6757 -> gap -3.12 pts FAIL Phi(5.5/13.5) = 0.6581 vs fair ML 0.6757 -> gap -1.75 pts PASS Phi(6.0/13.5) = 0.6716 vs fair ML 0.6757 -> gap -0.40 pts PASS implied sigma_margin if the ML is right: at S=5.5 -> 12.07 | at S=6.0 -> 13.17 (spec 13.5) ```

The market's own median posted number (5.5) is less coherent with its own moneyline than the highest posted number (6.0). At 6.0 the implied sigma is 13.17 against the spec's 13.5 — the closest agreement in four runs. The anchor is nonetheless set at 5.5, because the spec instructs anchoring to consensus rather than to whichever number the model prefers; the 6.0 preference is printed rather than acted on, and the coherence test below covers the whole band.

⚠️ BLOCK C MOVEMENT: NOT ESTABLISHABLE. Four same-day sources give four different stories (see Discards). FOX and SportsBettingDime report the identical pair of numbers in opposite order. Two sources both claiming to quote DraftKings today disagree on DraftKings' own number (−4.5 vs −5.5).

The one robust claim: the total has fallen roughly 2.5–3.5 points, from somewhere in 46.5–48.5 down to 44–44.5. Every source agrees on that, and thespread's reading is the most plausible mechanism: *"Totals that fall that far in the offseason usually reflect changed personnel rather than late money."* A.J. Brown's departure, Barkley's YPC collapse from 5.8 to 4.1, Philadelphia's ground game falling by more than a thousand yards, and new offensive coordinators on both sidelines are exactly that kind of repricing.

⚠️ RLM = NOT DETERMINABLE. The spec's definition requires knowing which way the line moved. It is not knowable from the sources opened. No RLM claim is made in either direction. This is a deliberate refusal, not an oversight: last run the same test produced a clean strong-form RLM finding on BAL @ IND because the movement was unambiguous. Here it is not, and asserting one would be manufacturing a signal out of a sourcing failure.

ANCHORING COHERENCE TEST — run across the whole four-number band:

``` L1 anchor 5.5/44.5, grade 5.5/44.5 : fav 0.4987 dog 0.5013 push 0.0000 over 0.5019 L2 anchor 5.5/44.5, grade 4.5/44.0 : fav 0.5287 dog 0.4713 push 0.0000 over 0.5019 L3 anchor 4.5/44.0, grade 4.5/44.0 : fav 0.4991 dog 0.5009 push 0.0000 over 0.4810 L2b anchor 5.5/44.5, grade 6.0/44.5 : fav 0.4688 dog 0.5013 push 0.0298 L3b anchor 6.0/44.5, grade 6.0/44.5 : fav 0.4842 dog 0.4859 push 0.0299

L2 vs L3 dog-cover divergence (4.5 branch) : -2.96 pts L2b vs L3b dog-cover divergence (6.0 branch): +1.55 pts properly-anchored dog-cover swing, 4.5 -> 6.0: +1.50 pts ```

The 4.5 branch produces the largest anchor-driven divergence measured in four runs: 2.96 points. Anchoring at consensus 5.5 and grading Washington +4.5 gives a dog cover of 47.13%; anchoring properly at 4.5 gives 50.09%. That three-point gap is pure anchor artifact. Any Washington +4.5 play is declined on the coherence test, before tier. Fourth consecutive run in which this gate has killed something.

---

[7] WEATHER LINE

WVI = 0.0 | rationale: Lincoln Financial Field is outdoor, so every threshold was evaluated rather than short-circuited. FOX Weather (game page, fetched today) shows Philadelphia at 74 °F with wind 3 mph ESE currently and 83 °F across the 7:00–10:00 block; ESPN's header carries 82 °F. No sustained wind ≥15 mph, no gusts ≥25, no precipitation ≥60% at kickoff, temperature far above 20 °F and below the 85 °F heat threshold. No trigger fires.

*Stated rather than buried: 83 °F sits within two degrees of the heat note. That branch contributes 0 to the total by rule in any case, so the conclusion is not sensitive to it.* Zero applied to μ_total, once, in §8.

---

[8] HOME-TEAM INFO LINE

Home team: Philadelphia Eagles — HFA already reflected in the market line; no model adjustment applied. (Market HFA ≈ 1.5–2.0 pts, true ≈ 2.0–2.5; additions banned under Part 6.)

---

[9] MSI / DOG-VALUE

Market Disagreement Module: NOT TRIGGERED.

Dog ML EDGE = +1.75% (or +2.62% excluding the suspected-stale quote). Both are below the 6% trigger and far below the 10% audit-alone threshold. Corroborators cannot be evaluated:

• RLM toward the dog: NOT DETERMINABLE — Block C is contradictory (§6).
• Money% exceeds bets% on the dog by ≥15 pts: INVERTED and stale. The only split available is five days old and predates both injury reports: Washington had 31% of tickets and 8% of handle. Money on the dog sat 23 points below tickets — the opposite of the DOG VALUE signature — but the reading cannot be trusted as current.

Neither corroborator is usable, and the primary gate does not open regardless. The module is correctly silent.

Underdog checklist: X/7 → N/A — FORCED. The seven NFL underdog criteria have never been enumerated anywhere in v3.1. Eleventh consecutive run in which this field cannot be scored.

---

[10] FINAL SCORE PREDICTION

Philadelphia 24 — Washington 20 *(presentation only, snapped to the Appendix B realistic grid)*

Sum 44 vs μ_total 44.5 (Δ −0.5, inside ±2). Margin 4 vs μ_margin 5.5 (Δ −1.5, inside ±2). 24-20 is on the Appendix B grid and satisfies C2.

*This scopes N-8 correctly.* Last run at μ_total 48.0 no grid pair could satisfy C2, because the grid jumps from sum 44 to sum 51. At 44.5 the grid works fine. The hole is specifically sums 45–50, not the grid in general — the ledger item is narrowed accordingly. No probability in this report is derived from this score.

---

[11] TRUE SCORE SIMULATION

Monte Carlo, NumPy `default_rng(7)`, N = 100,000. PHASE = REGULAR, RCF = False, μ_margin = 5.5, μ_total = 44.5, σ_margin = 13.5, σ_total = 10.0. Ties resolved by the 50/50 OT coin flip.

``` P(fav win) 0.658 | P(dog win) 0.342 | P(tie) 0.000 P(fav covers 5.5) 0.499 (push 0.000) P(over 44.5) 0.502 (push 0.000) Mean score fav 25.0 — dog 19.6 Most frequent scores: ['23-20 (0.26%)', '23-19 (0.25%)', '25-20 (0.25%)', '22-19 (0.24%)', '26-17 (0.24%)'] ```

CLAMP DIAGNOSTIC — run via `clamp_diagnostic.py`, not hand-rolled (F-40):

``` inputs spread 5.5 total 44.5 wvi 0.0 sd_m 13.5 sd_t 10.0 implied dog team total : 19.50 P(dog clamped to exactly 0) : 1.15% mu_total bias from floor : +0.062 pts P(over) unclamped / clamped : 0.5016 / 0.5019 P(over) shift from floor : +0.02 pts shutouts among top-5 modes : 0/5 ```

Measurements cited, VERDICT line not cited (F-34). Governing z-distance (F-40): `z = 19.50 / 8.40 = 2.321`.

N-10 now has four points and the monotone relationship is clean:

| Run | z | zero-mass | μ_total bias | P(over) shift | shutouts in top-5 | |---|---|---|---|---|---| | NE @ CLE (preseason, T 35.5) | 1.06 | 5.51% | +0.229 | +0.51 pts | 2/5 | | ATL @ PIT (T 41.5) | 2.11 | 1.93% | +0.090 | +0.05 pts | 0/5 | | WAS @ PHI (T 44.5) | 2.32 | 1.15% | +0.062 | +0.02 pts | 0/5 | | BAL @ IND (T 48.0) | 2.65 | 0.47% | +0.046 | +0.00 pts | 0/5 |

Four measurements, strictly monotone in z, with a sharp elbow below z ≈ 2. The proposed ratification in N-10 stands and is strengthened: z ≥ 2.0 → negligible, cite and move on; z < 1.5 → correction mandatory and top-5 modes must be cleaned.

---

[12] FIRST TEAM TO SCORE

P(PHI) = 0.5 + 0.4 × (0.6757 − 0.5) = 0.570 · P(WAS) = 0.430 REGULAR phase — no preseason clamp. Not priced on any retrieved board → no verdict.

---

[13] TOTAL O/U — with CF%

| Market | Model | Vig-free bar | EDGE | Verdict | |---|---|---|---|---| | Over 44.5 (−112, FanDuel) | 50.19% | 50.43% | −0.25% | No edge | | Under 44.5 (−108, FanDuel) | 49.81% | 49.57% | +0.25% | No edge | | Over 44.0 (−105, thespread) | 50.19% | 51.08% | −0.89% | No edge | | Under 44.0 (−115, thespread) | 47.91% *(2.30% push at 44)* | 48.92% | −1.01% | No edge |

Clamp artifact on P(over) is +0.02 pts — immaterial.

The total is the tightest market on the board: a quarter of a percentage point. That is what happens when the market has already done a 3-point repricing and the model is anchored to the result of it.

C-5 · SEASON-OPENER TRACKING LINE (record, do not adjust): > Week 1 opener. Market total 44.0–44.5 (anchor 44.5) · model total 44.5 · actual: [to be recorded post-game]. This is the third opener logged today (ATL/PIT 41.5, BAL/IND 48.0, WAS/PHI 44.5). C-5 stands at n=2 NFL from the original audit; adding three today takes the series to n=5 once results land. No point adjustment applied. Revisit at n ≥ 10.

---

[14] SPREAD — with EDGE

| Market | Model | Vig-free bar | EDGE | Verdict | |---|---|---|---|---| | PHI −5.5 (−105, FanDuel) | 49.87% | 48.92% | +0.94% | No edge | | PHI −5.5 (−108, DK via SI) | 49.87% | 49.57% | +0.30% | No edge | | WAS +5.5 (−115, FanDuel) | 50.13% | 51.08% | −0.94% | No edge | | WAS +5.5 (−112, DK via SI) | 50.13% | 50.43% | −0.30% | No edge | | PHI −5.0 (−115, thespread) | 49.87% | 51.08% | −1.21% | No edge | | WAS +5.0 (−105, thespread) | 50.13% | 48.92% | +1.21% | No edge | | WAS +4.5 (DK) | 50.09% *properly anchored* | price not published | — | DECLINED — 2.96-pt anchor divergence (§6) | | *push at exactly 6, disputed branch* | 2.98% | — | — | — |

KEY-NUMBER ZONE: ACTIVE, and the whole live band sits in it. μ_margin = S = 5.5, exactly 0.50 pt from key number 6; the band's lower end (4.5) is 0.50 from key number 4. Both 4 and 6 are on the spec's second-tier key list (10, 4, 6, 14). Lean threshold raised from 3% to ≥5%. Nothing is within a quarter of it.

The 0.5-vs-0.25 zone-width conflict fires for the fourth consecutive run. At exactly 0.50 the zone fires under the written spec and not under the operating note. Stricter reading applied. Verdict unaffected — every edge is under 1.3%.

MONEYLINE:

| | Model | Vig-free bar | EDGE | |---|---|---|---| | PHI −228 to −250 | 65.82% | 67.57% | −1.75% | | WAS +180 to +205 | 34.18% | 32.43% | +1.75% | | *WAS, excluding suspected-stale quote* | 34.18% | 31.56% | +2.62% |

+2.62% is the largest number on this board and it is still below the 3% floor and half the 5% key-number threshold. Declined under both anchorings.

Sigma note (N-1 control): implied σ_margin is 12.07 at S=5.5 and 13.17 at S=6.0, against the spec's 13.5. At the anchored number the gap is 1.4 sigma-points, which is where the moneyline edge is coming from — the same mechanism that manufactured +2.90% on ATL @ PIT this morning, at about 60% of the magnitude. The ML "edge" here is substantially a sigma artifact and is declined on that basis as well as on tier.

---

[15] FIRST TD TABLE / TEAM

N/A (SF — no first-touchdown-scorer market located on any of the four boards retrieved; Gate 2 log below). The value exists in principle; retrieval was attempted across four sources and failed. Not SI.

Model: P(PHI scores first) = 57.0%. Not distributed across players, for the reason in §16.

---

[16] ANYTIME TD TABLE / TEAM

λ_team = model team total ÷ 10: PHI 2.500 · WAS 1.950 (÷9 alternative: 2.778 / 2.167).

A full eight-player anytime-TD board was retrieved (NBC Sports Philadelphia 9/11, BetMGM 9/9; best price per player taken, then the Part 10.3 single-sided 2.5-point haircut applied). This is the second consecutive run able to test the allocation against a complete market.

| Player | Market best ATD | Implied | After haircut | Model (spec role weights) | Model − market | Market-implied role weight | Spec weight | |---|---|---|---|---|---|---|---| | Saquon Barkley (PHI) | −120 | 54.6% | 52.0% | 44.2% | −7.8 | 1.76 | 1.4 | | Jalen Hurts (PHI) | +125 | 44.4% | 41.9% | 11.8% | −30.2 | 1.31 | 0.3 | | DeVonta Smith (PHI) | +175 | 36.4% | 33.9% | 36.8% | +2.9 | 0.99 | 1.1 | | Dallas Goedert (PHI) | +210 | 32.3% | 29.8% | 34.1% | +4.3 | 0.85 | 1.0 | | J. Croskey-Merritt (WAS) | +150 | 40.0% | 37.5% | 36.6% | −0.9 | 1.45 | 1.4 | | Rachaad White (WAS) | +230 | 30.3% | 27.8% | 22.9% | −4.9 | 1.00 | 0.8 | | Jayden Daniels (WAS) | +260 | 27.8% | 25.3% | 9.3% | −16.0 | 0.90 | 0.3 | | Terry McLaurin (WAS) | +270 | 27.0% | 24.5% | 30.1% | +5.5 | 0.87 | 1.1 | | *Wicks, RB2, WR3 (PHI); Diggs, Okonkwo, Burks (WAS)* | no line retrieved | — | — | 34.1 / 28.3 / 15.4 / 27.7 / 27.7 / 12.2% | — | — | — |

⚠️ N-2 CONFIRMED OUT-OF-SAMPLE, AND THE FIX IS NOW QUANTIFIED

Last run I retracted the "uniform inflation" half of N-2 and substituted a shape diagnosis: the role weights are too flat, under-pricing goal-line backs and rushing quarterbacks and over-pricing the WR/TE tail. This game was not used to form that hypothesis and it confirms every element of it:

• Jalen Hurts: model 11.8% vs market 41.9% — a 30.2-point miss and a 3.6× understatement. This is the single largest model error recorded in four runs, and it is caused entirely by the "rushing QB 0.3" constant being applied to the most prolific short-yardage rushing quarterback in football.
• Jayden Daniels: model 9.3% vs market 25.3% — a 16.0-point miss. Second rushing QB, same direction.
• Saquon Barkley: −7.8, market-implied weight 1.76 against the spec's 1.4 for a goal-line back.
• Croskey-Merritt: −0.9, implied weight 1.45 — essentially exact. He is a lead back but *not* a dominant goal-line back like Henry, Taylor or Barkley, and the 1.4 weight fits him properly. That is the control case: the weight is not wrong everywhere, it is wrong for specific archetypes.
• WR1/TE1 tail over-priced by 2.9 to 5.5 points (Smith, Goedert, McLaurin), implied weights 0.85–0.99 against spec 1.0–1.1.

Market-implied weights, pooled across the two games where full boards existed:

| Archetype | Spec weight | Market-implied | Cases | |---|---|---|---| | Rushing QB | 0.3 | 0.90 – 1.31 | Hurts, Daniels (+ Lamar Jackson −11.7 pts last run) | | Dominant goal-line RB | 1.4 | ~1.76 | Barkley (+ Henry −12.2, Taylor −16.5 last run) | | Ordinary lead RB | 1.4 | ~1.45 | Croskey-Merritt | | RB2 | 0.8 | ~1.00 | White | | WR1 | 1.1 | 0.87 – 0.99 | Smith, McLaurin | | TE1 | 1.0 | ~0.85 | Goedert |

Proposed amendment, now backed by measurements rather than reasoning: raise rushing QB from 0.3 to ~1.0 (ordinary) / ~1.4 (elite designed-run), raise a dominant goal-line back to ~1.8, raise RB2 to ~1.0, and lower WR1 to ~1.0 and TE1 to ~0.85.

Unallocated residual, now measured across four teams: 26.9% / 26.0% (last run) and 18.2% (PHI) / 29.8% (WAS) this run — mean 25.2%. The proposal to hold ~25% of λ_team unallocated and to make C4 test the residual rather than the sum survives a second game.

No anytime-TD verdict is issued on any player. The two largest apparent "edges" on this board are Hurts and Daniels, where the model is 30 and 16 points *below* the market — precisely the archetype the allocation is worst at. C4 passes arithmetically and is qualified in §25.

---

[17] PLAYER BREAKOUT (BRI)

N/A (SI — insufficient current-season sample).

1. Structural fact: BRI requires ≥3 current-season games; zero regular-season games have been played by either club. 2. Gate 0 confirms Block I's primary was fetched (FOX all-32 2025 tables + 2025 player leaders). No unopened source is being masked. 3. What would have to change: games would have to be played. SI stands.

---

[18] PLAYER PROJECTION TABLES — passing, rushing, receiving

GATE 3 — RATE-TIER WALK, with the mandatory C-1 column

C-1 factors, from all 32 teams' 2025 final defensive rates:

| Direction | Pass factor | Rush factor | |---|---|---| | WAS offense vs PHI defense | 0.905 (189.8 ÷ 209.70; PHI 8th vs pass) | 1.064 (124.4 ÷ 116.94; PHI 22nd vs run) | | PHI offense vs WAS defense | 1.156 (242.5 ÷ 209.70; WAS 28th vs pass) | 1.213 (141.8 ÷ 116.94; WAS 30th vs run) |

This is the most asymmetric C-1 pair of the four runs — Philadelphia is boosted in both phases, Washington is cut through the air and boosted only modestly on the ground. That makes it the cleanest available test of the N-7 diagnosis, and the test is run explicitly below.

| Player | Tier | Source | Rate used | C-1 factor | Adjusted | Volume basis | |---|---|---|---|---|---|---| | Jalen Hurts (PHI) | T1 | 2025: 3,224 yds / 454 att | 7.101 Y/A | 1.156 | 8.212 | posted market line — 28.5 att | | Saquon Barkley (PHI) | T1 | 2025: 1,140 / 280 att | 4.071 YPC | 1.213 | 4.937 | posted market line — 16.5 att | | DeVonta Smith (PHI) | T1 | 2025: 1,008 / 77 rec | 13.091 yds/rec | 1.156 | 15.139 | posted market line — 5.5 rec | | Dallas Goedert (PHI) | T3 (la) | 2025 rate SF | 7.3 yds/tgt | 1.156 | 8.44 | posted market line — 3.5 rec | | Jayden Daniels (WAS) | T3 (la) — T1 BLOCKED | 188 att < the 200-att gate, by twelve | 6.2 Y/A | 0.905 | 5.612 | posted market line — 29.5 att | | Terry McLaurin (WAS) | T1 | 2025: 582 / 38 rec | 15.316 yds/rec | 0.905 | 13.863 | posted market line — 3.5 rec | | J. Croskey-Merritt (WAS) | T3 (la) | 2025 rate SF (rookie, partial) | 4.3 YPC | 1.064 | 4.574 | posted market line — 11.5 att | | Stefon Diggs (WAS) | T3 (la) | prior-team 2025 rate SF | 7.3 yds/tgt | 0.905 | 6.61 | posted market line — 3.5 rec | | Rachaad White (WAS) | T3 (la) | prior-team 2025 rate SF | 4.3 YPC | 1.064 | 4.574 | posted market line — 6.5 att |

⚠️ Jayden Daniels' tier-1 rate is blocked by twelve attempts. He threw for 1,262 yards at 6.7 Y/A across seven injury-plagued 2025 appearances — 188 attempts against Part 10.2's 200-attempt gate. The spec sends him to the league-average constant. Both branches are printed below; the gate is arguably doing harm here, since 188 attempts of a known quarterback's own rate is better evidence than a league constant. Logged.

PASSING PROJECTIONS

| QB | Line (book, age) | Model | Gap | Verdict | |---|---|---|---|---| | Jalen Hurts | 211.5 (−114/−114) phlsportsnow today · 214.5 NBC 9/11 · 209.5 SBD 9/7 · 1.5 pass TD (+114/−152) · 28.5 att | T1: 234.0 · *(la): 204.3* | +10.7% · *−3.4%* | NO VERDICT | | | *market-implied Y/A = 211.5 / 28.5 = 7.42* | | | | | Jayden Daniels | 205.5 NBC 9/11 · 201.5 SBD 9/7 · 0.5 pass TD (−240/+174) · 29.5 att | (la): 165.5 · *T1-if-permitted: 178.9* | −19.4% · *−12.9%* | NO VERDICT | | | *market-implied Y/A = 205.5 / 29.5 = 6.97* | | | |

Daniels carries a 0.5 passing-touchdown line — the market is pricing better-than-even odds that he throws zero touchdown passes against Philadelphia's 8th-ranked pass defense. That is one of the lowest passing-TD lines on the entire Week 1 board.

RUSHING PROJECTIONS

| Player | Line (book, age) | Model | Gap | Verdict | |---|---|---|---|---| | Saquon Barkley rush | 75.5 · 16.5 att NBC 9/11 · 76.5 (−115/−112) phlsportsnow today · 77.5 · 18.5 att SBD 9/7 | T1: 81.5 · *(la): 86.0* | +6.5% · *+12.5%* | NO VERDICT | | Barkley rush + rec | 98.5 (−115/−111) SBD 9/7 | T1: 100.1 | +1.6% | NO VERDICT | | J. Croskey-Merritt rush | 40.5 · 11.5 att NBC 9/11 · 42.5 · 10.5 att SBD 9/7 | (la): 52.6 | +29.9% | NO VERDICT | | Jayden Daniels rush | 36.5 · 8.5 att NBC 9/11 · 38.5 (−120/−111) BetMGM | 2025 rush rate SF | — | — | | Jalen Hurts rush | 26.5 (−120/−111) BetMGM · FOX FACTS: 5.6 rush att/g, under 7.5 att in 11 of last 12 | 2025 rush-yard rate SF | — | — | | Rachaad White rush | 23.5 · 6.5 att SBD 9/7 · 42.5 rush+rec | (la): 29.7 | +26.5% | NO VERDITC → NO VERDICT |

Barkley's attempt line fell from 18.5 (Sept 7) to 16.5 (Sept 11) — two full carries. At the newer volume his tier-1-adjusted projection lands within 6.5% of the yardage line and his rush+rec projection within 1.6%. That is the closest agreement between model and market anywhere in this run.

RECEIVING PROJECTIONS

| Receiver | Line (book, age) | Model | Gap | Verdict | |---|---|---|---|---| | DeVonta Smith | 66.5 · 5.5 rec NBC 9/11 · 67.5 (−113/−113) phlsportsnow today · 64.5 BetMGM · 65.5 SBD | T1: 83.3 | +25.2% | NO VERDICT | | Terry McLaurin | 52.5 · 3.5 rec NBC 9/11 · 51.5 BetMGM/SBD | T1: 48.5 | −7.6% | NO VERDICT | | Stefon Diggs | 42.5 · 3.5 rec NBC 9/11 · 41.5 (−120/−111) BetMGM | (la): 35.6 | −16.2% | NO VERDICT | | Dallas Goedert | 34.5 · 3.5 rec NBC 9/11 · 35.5 BetMGM | (la): 45.5 | +31.9% | NO VERDICT | | Dontayvion Wicks (PHI) | 27.5 (−120/−110) BetMGM | rate SF | — | — | | J. Croskey-Merritt rec | FOX FACTS: under 7.5 rec yds in 12 straight — 1.2 rec yds/g | — | — | context only |

⚠️ N-7 OUT-OF-SAMPLE TEST — 6 OF 6, THE DIAGNOSIS HOLDS

Last run I diagnosed the passing overshoot as a volume double-count inside C-1: the factor is built from per-game yardage allowed and then applied to a per-attempt rate that is multiplied by an attempt count, so volume enters twice. That predicts a specific, falsifiable thing: the sign of the passing-projection error should follow the sign of (C-1 pass factor − 1), regardless of which tier the rate came from.

This game supplied the strongest available test, because the two pass factors point in opposite directions — 1.156 for Philadelphia, 0.905 for Washington.

| Run | QB | C-1 pass factor | (factor − 1) | Projection error | Sign | |---|---|---|---|---|---| | ATL @ PIT | Cooper Rush | 1.163 | +0.163 | +21.8% | MATCH | | ATL @ PIT | Aaron Rodgers | 0.956 | −0.044 | −13.4% | MATCH | | BAL @ IND | Lamar Jackson | 1.182 | +0.182 | +20.5% | MATCH | | BAL @ IND | Daniel Jones | 1.182 | +0.182 | +2.0% | MATCH | | WAS @ PHI | Jalen Hurts | 1.156 | +0.156 | +10.7% | MATCH | | WAS @ PHI | Jayden Daniels | 0.905 | −0.095 | −12.9% | MATCH |

Six for six across three games and six quarterbacks, including both signs, and including one case (Daniels) where the rate came from the league-average constant rather than tier-1. That last point is the decisive one: if the error were a tier-1 problem, an `(la)`-sourced projection would not err in the factor's direction. It does. The defect is in C-1, not in the rate tiers.

N-7's fix is unchanged and now well-supported: build C-1 from per-attempt allowed rates, not per-game yardage.

C-4 · REGULAR-PHASE QB-CHANGE CONDITION, stated as required: both passing projections carry an implicit "if he plays the whole game." Jayden Daniels appeared in only seven games in 2025 and was shut down for the season, and he missed both Eagles meetings last year; Marcus Mariota is the backup and was a full participant all week. Prefer the team-level passing read over the per-QB line when grading the total.

Script Note (Part 10.7): NOT triggered. Favourite win probability 65.82% (< 78%) and |μ_margin| = 5.5 (< 10). No sigma widening on any prop.

C-3 · TURNOVER TERM — STILL NOT EVALUABLE, FOURTH CONSECUTIVE RUN, BUT THE CLOSEST YET. Washington finished 31st in turnover differential at −13 in 2025. The C-3 trigger, however, is *"either team bottom-10 in giveaways or the opponent top-10 in takeaways"* — differential rank does not establish either component, and beat coverage frames Washington's problem as a need for takeaways, which is the half that does not trigger the rule. The sigma widening is not applied. Four runs, four failures to feed the engine's self-declared highest-priority structural gap. Escalated again.

---

[19] EDGE SUMMARY

• WAS ML +180/+205: EDGE +1.75% (+2.62% excluding the suspected-stale quote) → No edge, and substantially a sigma artifact
• PHI ML −228/−250: EDGE −1.75% → No edge
• WAS +5.0 (−105, thespread): EDGE +1.21% → No edge
• PHI −5.5 (−105, FanDuel): EDGE +0.94% → No edge
• PHI −5.5 (−108, DK): +0.30% → No edge
• PHI −5.0 (−115): −1.21% → No edge
• WAS +4.5 (DK): DECLINED — 2.96-pt anchoring divergence, largest of four runs
• Under 44.5 (−108): EDGE +0.25% → No edge
• Over 44.5 (−112): EDGE −0.25% → No edge
• All 14 player props: priced, projected, no verdict
• All 8 anytime TDs: no verdict — allocation shape demonstrably wrong, worst on the two quarterbacks
• First TD: N/A (SF)

Largest edge anywhere: +2.62%, under the most favourable anchoring, against a key-number threshold of 5%.

---

[20] INSIGHTS

A big quality gap, a big total drop, and a market whose direction of travel cannot be read.

The 2025 tables in §3 are as lopsided as any matchup run this session. Philadelphia was 5th in scoring defense and 1st in the NFL in red-zone touchdown rate; Washington finished last in total defense and 31st in red-zone defense. Philadelphia won 11 games and reached the Wild Card round; Washington won five. And yet the market is only laying 4.5 to 6 points, and the total has been cut by roughly three.

What the three-point total drop is actually about. It is not late money — every source that dates it puts the move in the offseason. A.J. Brown left for New England, taking the other half of Philadelphia's receiving corps with him. Barkley's yards per carry fell from 5.8 to 4.1 and the Eagles' ground game lost more than a thousand yards year over year. Both teams replaced their offensive coordinator — Sean Mannion in Philadelphia, David Blough in Washington — and Washington replaced its defensive coordinator too. Four new coordinators and a gutted receiving room is exactly the kind of thing that moves a total three points and a spread not at all.

Strengths, Philadelphia. One of the best offensive lines in football, an interior defense of Jalen Carter and Jordan Davis, and the league's best red-zone offense meeting the league's 31st-ranked red-zone defense. Jalen Hurts threw a career-high 25 touchdowns against six interceptions with a 98.5 rating. Nolan Smith and Jalyx Hunt face what one Philadelphia outlet called "an entirely exploitable Commanders offensive line."

Weaknesses, Philadelphia. Jonathan Greenard, the headline defensive addition, is out. Brandon Graham is retired. The offense is installing a new scheme in Week 1 with DeVonta Smith drawing coverage that used to be split with A.J. Brown. They went 0-3 in the preseason and were outscored 41-84.

Strengths, Washington. Completely healthy — Dan Quinn cleared all 53, including both of his best pass rushers in Chaisson and Oweh, who between them out-pressured any single Commander in 2025. The 4th-ranked rushing offense meets Philadelphia's 22nd-ranked run defense, which is the one place the matchup tilts their way. Jayden Daniels is back after missing seven games and both Eagles meetings last year. Stefon Diggs joins Terry McLaurin. They beat Philadelphia in the most recent meeting.

Weaknesses, Washington. Last in total defense. 30th against the run against a team that will run. 31st in turnover differential at −13. A quarterback whose own passing-touchdown line is set at 0.5. Three new faces in the secondary against a first-place offense.

Where the edge actually is: nowhere at these prices. The total is priced within a quarter of a percentage point of the model. The spread's four live numbers straddle two key numbers and the model sits inside them. The largest number on the board is a moneyline gap that the sigma analysis attributes mostly to the spec's own fixed variance rather than to anything about this football game.

---

[21] BOOK EXPOSURE READ

Insufficient current data — and this section is qualified more heavily than in any prior run.

The only splits available are five days old and predate both final injury reports. As of 9/8, Philadelphia held 69% of spread tickets and 92% of spread handle, and 89% of moneyline handle as of 9/5. The Under held 55% of tickets and 85% of handle. That is an overwhelming large-money favourite profile.

What cannot be said, and is therefore not said: whether the line moved with that money or against it. Four same-day sources give four incompatible answers, two of them attributing contradictory numbers to the same book (§6). Without a reliable direction, a 92%-handle figure is not a signal — it is just exposure. A 69/92 split moving *with* the line is an ordinary steamed favourite; the same split moving *against* it would be one of the strongest reverse-line-movement readings of the season. The engine cannot distinguish those two worlds from the sources it opened, and declines to guess.

No handle figures are estimated and no movement direction is asserted.

---

[22] PARLAY (honest)

Step 1 — best five legs by model probability (no leg qualifies on EDGE, so highest-probability legs are used, per Part 12):

| # | Leg | Model prob | |---|---|---| | 1 | Philadelphia ML | 0.6582 | | 2 | Philadelphia first team to score | 0.5703 | | 3 | Over 44.5 | 0.5019 | | 4 | Washington +5.5 | 0.5013 | | 5 | Saquon Barkley anytime TD | 0.4420 |

True joint probability = 0.6582 × 0.5703 × 0.5019 × 0.5013 × 0.4420 = 4.18%.

Independence caveat, mandatory: all five legs come from one game and are strongly correlated. Legs 1 and 2 move together; leg 4 directly opposes leg 1; leg 5 is coupled to both the total and the game script. Multiplying them as independent is wrong and the true figure is not knowable from this board. Leg 5 is additionally drawn from the anytime-TD allocation that §16 shows to be mis-shaped — and for a dominant goal-line back specifically, too low (model 44.2% against a market 52.0%), so that leg is understated while the product overall is overstated.

Step 2 — the 99% arithmetic, plainly. A five-leg parlay reaching 99% requires 0.99^(1/5) ≈ 99.80% per leg. The highest model probability anywhere on this board is 65.82%. No real five-leg parlay reaches 99%, and none reaches 99.99%. No parlay is recommended and no confidence figure is printed, because none can be computed honestly from this board.

---

[23] RAW PREDICTION — true probabilities, nothing inflated

• Moneyline: PHI 65.82% · WAS 34.18% · tie 0.00% (OT resolved)
• Spread 5.5: PHI covers 49.87% · WAS covers 50.13% · push 0.00%
• Spread 4.5 (disputed branch, properly anchored): PHI 49.91% · WAS 50.09%
• Spread 6.0 (disputed branch, properly anchored): PHI 48.42% · WAS 48.59% · push 2.99%
• Total 44.5: Over 50.19% · Under 49.81% · push 0.00%
• First team to score: PHI 57.0% · WAS 43.0%
• Model team totals: PHI 25.00 · WAS 19.50 · λ_team 2.500 / 1.950
• Top-5 simulated finals: 23-20 (0.26%) · 23-19 (0.25%) · 25-20 (0.25%) · 22-19 (0.24%) · 26-17 (0.24%)
• Passing yards: Hurts 234.0 T1 / 204.3 (la) — market 211.5 · Daniels 165.5 (la) / 178.9 T1-blocked — market 205.5
• Rushing yards: Barkley 81.5 (market 76.5) · Croskey-Merritt 52.6 (market 40.5) · White 29.7 (market 23.5) · Hurts and Daniels rushing N/A (SF)
• Receiving yards: DeVonta Smith 83.3 (market 66.5) · McLaurin 48.5 (market 52.5) · Diggs 35.6 (market 42.5) · Goedert 45.5 (market 34.5) · Wicks N/A (SF)
• Rush + receiving: Barkley 100.1 (market 98.5)
• Anytime TD: see §16 — mis-shaped, no verdicts
• Alt spreads / alt totals: N/A (SF) — no priced alternate ladder retrieved, and the base spread is unresolved across four numbers

Every leg above is below the legacy 80% display floor. True probabilities are shown; none inflated.

---

[24] TOP PLAYS

NO QUALIFYING EDGE.

No spread, moneyline, total, prop or touchdown market on this game clears its EDGE tier. The largest figure on the board is +2.62% on the Washington moneyline under the most favourable of two defensible anchorings, against a key-number threshold of 5% — and the sigma analysis attributes most of that gap to the spec's fixed σ_margin of 13.5 against a market-implied 12.07, not to any view about the football. Washington +4.5 was killed by the anchoring coherence test at a 2.96-point divergence before it reached this section.

There is no market signal to record this time either. Block C is contradictory enough that the reverse-line-movement question cannot be answered, and Block D is five days stale. That is a sourcing failure, not a finding, and it is recorded as one.

> ### C-6 · MARKET-DEPENDENCY NOTE (mandatory, ledger empty) > The ledger is empty, so every distributional forecast in this run is the market's forecast. μ_margin is the market's spread and μ_total is the market's total. A no-bet verdict means *no disagreement with the market was found* — it does not mean the forecast is likely accurate. On 2026-08-27 the market was 14.5 points wrong on the total and this engine was wrong by exactly the same amount, because it had nothing else to be. This is now eight consecutive runs with an empty ledger, and four in a single session.

⚠️ C10 UNRESOLVED. The spec requires that "if no qualifying edge, the no-bet line prints verbatim." No verbatim no-bet string is defined anywhere in v3.1. The wording above is the model's own canonical phrasing from Global Rules and Part 11; it is not a defined verbatim string and is not presented as one. C10 cannot be marked PASS. Fourth consecutive run in which this fires.

---

[25] COHERENCE CHECK C1–C13

| # | Check | Result | |---|---|---| | C1 | Winner = higher win prob = higher model total | ✅ PHI 65.82%, model total 25.00 | | C2 | Headline sum within ±2 of μ_total; diff within ±2 of μ_margin | ✅ 24-20 → sum 44 (Δ −0.5), diff 4 (Δ −1.5). On the Appendix B grid — N-8's hole is scoped to sums 45–50 | | C3 | Every verdict matches its EDGE tier and clamps | ✅ all No edge; key-number 5% threshold applied; no Lean cap | | C4 | Σ player TD λ within 5% of λ_team | ⚠️ Arithmetically PASS — QUALIFIED. Shape is wrong: rushing QBs under-priced by 16–30 points. See §16 | | C5 | Every named player passes roster check | ⚠️ DEGRADED (WAS clean / PHI partial) — Washington's official depth chart retrieved; Philadelphia personnel named from official reports and prop boards | | C6 | Weather appears exactly once | ✅ §7 only; outdoor thresholds evaluated; WVI 0.0 fed to μ_total once | | C7 | Ledger entries and nets within caps | ✅ ledger empty — 0.0 both teams, 0.0 to total | | C8 | Output numbers match sim output exactly | ✅ `sim_phi.py` run verbatim; `derive_phi.py` reproduces `default_rng(7)` seed and draw order identically | | C9 | Every missing block has N/A + code, all consequences applied | ✅ C (contradictory), D (stale), F (PHI side), C-3, and the SF items in §4/§15/§18 coded and wired | | C10 | No-bet line prints verbatim | ❌ CANNOT PASS — verbatim string undefined in v3.1 | | C11 | Date guard passed; no wrong-slate data | ✅ UTC artifact overridden; 5 discards/conflicts logged | | C12 | Gate 0 table printed; no SI on an unfetched block; breaker walked | ✅ table printed; the one SI (§17, BRI) passes all three tests; breaker tripped at 6 and walked with reasons | | C13 | Ledger-banned but retrievable prior data printed as tagged context | ✅ §3 2025 tables, §4 preseason and tier-3 rows, §4 five-meeting H2H — all season- and phase-tagged with the ban stated separately |

Gate result: 10 PASS · 2 QUALIFIED · 1 CANNOT-PASS. Released with defects on the face of the report.

Gate 1 player-module circuit breaker: N/A sections = 2 (first TD, BRI). At the threshold, not over.

GATE 2 — SEARCH LOGS

BLOCK L (props) — all four steps run:

| # | Query | Results | Adopted | Discarded (reason) | |---|---|---|---|---| | 1 | "Commanders Eagles Week 1 player props Barkley / Hurts / Daniels / McLaurin / Smith + 2025 stats" | 10 | NBC Sports Philadelphia full board 9/11 (pass/rush/rec/ATD), BetMGM full board 9/9, phlsportsnow today, SBD 9/7 | CBS Week-16 2025 Hurts/Mariota prop article — wrong season | | 2 | 2025 player-production pass (same query) | 10 | PhillyVoice 2025 leaders (Hurts, Barkley, Smith, Daniels, McLaurin), Yahoo 10-stats | — | | 3 | Book-affiliated game-article check | — | NBC Sports Philadelphia and BetMGM are both book- or network-affiliated boards | — | | 4 | Re-extract already-fetched pages — FOX game page FOX FACTS module | hit | Hurts 5.6 rush att/g (under 7.5 in 11 of 12); Croskey-Merritt 1.2 rec yds/g (under 7.5 in 12 straight); over 7-of-8 in the series | — | | — | first-TD-scorer board | not found on any of four boards | — | §15 coded SF, not SI | | — | Goedert / Wicks / Diggs / White 2025 rates | not returned | — | SF → (la) |

BLOCK D (splits) — all four steps run, and it still came up stale:

| # | Query | Results | Adopted | Discarded (reason) | |---|---|---|---|---| | 1 | "Commanders Eagles Week 1 odds spread total line movement betting splits today" | 10 | cleatz 9/8 bets+handle board | — | | 2 | "public betting percentage handle" (same pass) | — | cleatz 9/5 ML handle | — | | 3 | Splits-publisher check — SBD, cleatz, thespread, CBS/BetMGM | 4 publishers | cleatz only | SBD's Week-1 splits page carries no WAS/PHI entry; CBS/BetMGM article is prior-season | | 4 | Re-extract already-fetched pages | DK Network splits widget | — | server-renders MLB only — FIFTH confirmation | | — | post-injury-report split | none exists in any source opened | — | Block D coded 5 days STALE; DOG VALUE corroborator unusable |

BLOCK C (movement) — searched and found contradictory: four same-day sources, four incompatible stories, two of them attributing different numbers to DraftKings on the same day. Movement coded NOT ESTABLISHABLE; opener SF.

GATE 4 — BLOCK PROPAGATION

| Block | Status | Unlocked downstream | Degraded downstream | |---|---|---|---| | A | PASS | §1, §4, §5 corroboration, §7, §18 FOX FACTS anchors | — | | B | PASS (substitute: DK-attributed at two outlets, FOX module, FanDuel, thespread, ESPN) | §6 board, §13, §14 bars | no Lean cap; DK-vs-DK same-day conflict logged | | C | FAIL (contradictory) | — | RLM NOT DETERMINABLE (§6, §9, §21); opener SF | | D | PASS but 5 DAYS STALE | §21 Book Exposure with caveat | may not corroborate DOG VALUE (§9) | | E | PASS | §7 WVI 0.0 → §8.1 μ_total | — | | F | PARTIAL (WAS complete, PHI degraded) | §16 pool, §18 named set | PHI tier-1 rates for Goedert/Wicks → `(la)` | | G | PASS (published WAS depth chart + announced starters, per user directive) | §18 volume assumptions; §16 pool | "no announced plan" may not be cited as a blocker downstream | | H | PASS (official reports, both clubs, full participation grid on the WAS side) | §5, §20, ledger candidates | — | | I | PASS | §3, §18 C-1 factors, §2 | 2025 giveaway/takeaway ranks not retrieved → C-3 not evaluable | | J | PASS (five meetings) | §4 H2H context | — | | K | PASS (T1 SI → T2 full → T3 partial) | §4 | tier-3 ATS/O-U and the PHI Wild Card score SF | | L | PASS (substitute: four book-attributed boards) | §18 lines, §16 ATD board | staleness tagged per line |

No block is marked PASS and treated as FAIL anywhere downstream. Block C is the first block in four runs marked FAIL, and its consequence is propagated explicitly into three separate sections rather than quietly dropped.

---

# LEDGER — updated by this run

NEW

| # | Item | Status | |---|---|---| | N-13 | Block C has no integrity check, and four same-day sources can contradict each other completely. FOX and SportsBettingDime reported the identical pair of spread numbers in opposite order; FOX and SI both claimed to quote DraftKings today and gave different numbers (−4.5 vs −5.5). The spec treats line movement as a retrievable fact with a fallback chain, but provides no rule for what to do when the chain returns mutually exclusive answers. Fix: require ≥2 independent sources to agree on both the opening number and the direction before any movement or RLM claim is made; otherwise code the movement NOT ESTABLISHABLE, as done here. | open — high priority | | N-14 | F-45's freshness screen is worded one-directionally and should be widened. It was written because a stale quote *manufactured* edge. This run a suspected-stale quote (+180/−218, matching a 10-day-old capture) was suppressing edge — excluding it moved the dog moneyline from +1.75% to +2.62%. Fix: reword F-45 to require reporting the anchor with and without any quote that cannot be freshness-verified, in both directions. | open | | N-15 | Part 10.2's 200-attempt gate is too blunt at the margin. Jayden Daniels' 2025 tier-1 rate was blocked by twelve attempts (188 of 200), sending a known starter's own 6.7 Y/A to a league-average 6.2 constant and changing his projection from 178.9 to 165.5. Fix: allow a shrinkage band — between roughly 120 and 200 attempts, blend the player's own rate toward `(la)` in proportion to sample size rather than discarding it outright. | open |

CONFIRMED OUT-OF-SAMPLE THIS RUN

| # | Item | Status | |---|---|---| | N-7 | C-1's volume double-count — 6 of 6 sign matches across three games and six quarterbacks, including both factor directions and including one `(la)`-sourced projection. The `(la)` case (Daniels, −12.9% against a 0.905 factor) is decisive: if the error came from tier-1 rates, a league-average projection would not err in the factor's direction. It is C-1, not the rate tiers. Fix unchanged: build the factor from per-attempt allowed rates, not per-game yardage. | confirmed — highest priority | | N-2 | Shape diagnosis confirmed on a game not used to form it, and the fix is now quantified. Jalen Hurts model 11.8% vs market 41.9% (−30.2 pts, the largest model error in four runs); Jayden Daniels −16.0. Market-implied role weights: rushing QB 0.90–1.31 against the spec's 0.3; dominant goal-line RB ~1.76 vs 1.4; RB2 ~1.00 vs 0.8; WR1 0.87–0.99 vs 1.1; TE1 ~0.85 vs 1.0. Control case present: Croskey-Merritt, an ordinary lead back, comes in at 1.45 against the spec's 1.4 — the weights are wrong for specific archetypes, not everywhere. | confirmed, amendment drafted | | N-10 | Clamp-vs-z relationship now has four monotone measurements (z 1.06 → 5.51%; 2.11 → 1.93%; 2.32 → 1.15%; 2.65 → 0.47%) with a sharp elbow below z ≈ 2. Ready to ratify: z ≥ 2.0 negligible, z < 1.5 correction mandatory. Would retire F-34's unratified VERDICT thresholds. | ready to ratify | | N-8 | Narrowed. Appendix B satisfied C2 cleanly here at μ_total 44.5 (24-20, sum 44). The grid hole is specifically sums 45–50, not the grid generally. | narrowed | | N-1 | Third data point, consistent. Implied σ_margin 12.07 at the anchored 5.5 (13.17 at 6.0) against the spec's 13.5. Series: ATL@PIT 11.40 (phantom +2.90%), BAL@IND 12.56 (no artifact), WAS@PHI 12.07 (+1.75% to +2.62%, mostly artifact). The ML edge scales with the sigma gap, as predicted. | open, scoped | | N-5 | Circuit breaker tripped for the third consecutive run. A gate that fires on every well-sourced run is not discriminating. Fix: count only substitutions where no named secondary was used, or weight by block criticality. | open | | N-9 | Device clock non-monotonic across turns (04:26 → 04:43 forward here, backwards previously). Recorded as observed, never differenced. | open |

CARRIED FORWARD — fired again

| # | Item | Status | |---|---|---| | F-48 | Game-page injury modules vs official reports — fifth instance, new variant. FOX's INJURIES module listed Eli Stowers *Questionable* while its own PLAYER NEWS feed on the same page reported him placed on injured reserve Saturday. Previous instances had the module stale in the "healthy" direction; this one is stale in the "available" direction. Remedy unchanged: official report governs, module corroborates, and the player-news feed outranks the module when it carries a later transaction. | open (spec patch pending) | | F-34 / F-40 | Clamp thresholds unratified; script run, z printed (2.321). N-10 proposes ratification. | ready to ratify | | F-41 / C-1 | Opponent adjustment implemented in full; defective as specified — see N-7 | implemented, defective | | F-42 | Underdog checklist D7 "phase median total" undefined for REGULAR | open | | C-3 | No turnover term, NOT EVALUABLE for the fourth consecutive run — closest yet (WAS 31st in turnover differential, −13) but the rule keys on *giveaways* rank or opponent *takeaways* rank, neither of which was retrievable. Fix: name a specific ranked giveaway/takeaway source in the 3.3 source directory, or rewrite the trigger to key on turnover differential rank, which is widely published. | open — escalated, with a concrete fix | | C-5 | Opener total bias — three openers logged today (41.5 / 48.0 / 44.5), taking the series toward n≥5 once results land | tracking | | C-6 | Empty ledger = market-dependent forecast — note printed. Eight consecutive empty ledgers, four in this session alone. | ready to ratify | | — | P1 may be unsatisfiable as written. Eight for eight. Every candidate this run — Greenard out, Washington fully cleared, Stowers to IR, four new coordinators — was public before any datable market activity. The engine's data layer *is* the market's data layer, on the same Friday cycle. Either the ledger needs a class of input the market demonstrably lags on, or an empty ledger must be recognised as the expected result rather than an exception. | open — structural | | — | The 7 underdog criteria remain unenumerated. ELEVENTH consecutive forced N/A. | open — top blocker | | — | C10 verbatim no-bet string still undefined. Fourth consecutive fire. | open — top blocker | | — | DK Network splits and props tabs server-render to MLB. FIFTH confirmation. Appendix D's "confirmed C/D/L fallback" designation should be formally downgraded. | open | | — | Aggregator data-integrity failures — fourth instance in four runs. BUSR (inverted spread, wrong head coach); SI "Crazy Trends" (inverted sides); therx (phantom "Probable" designation); this run: FOX and SI both misquoting DraftKings on the same day, and a CBS splits article from a prior season surfacing as current. The named-entity sanity gate proposed on 8/27 should be extended to cover book attribution and season provenance. | open |

---

*Model: NFL Edge Engine v3.1 · Discipline: run-discipline v2, Gates 0–5 all printed · Calibration: calibration v1, C-1 through C-7 all applied or explicitly declined with reason · Simulation: NumPy default_rng(7), N = 100,000 · Scripts: market_math.py and clamp_diagnostic.py run from project knowledge, not reimplemented · Every figure dated and book-tagged; every N/A carries an SI, SF or SG code; nothing estimated, inflated or invented.*