# Footo Predicto 3.27.0 — Week 8 review and Week 9 forecasts

Updated November 4, 2026. Team results through November 2. All 32 rosters checked against current provider and official club evidence; medical coverage is partial early in the week. Saved forecasts are fixed before kickoff and are not rewritten from subsequent results.

## What Week 8 showed

The latest valid pregame forecasts picked **9 of 14 winners (64.3%)**. Mean absolute error was **9.07 points per team**, **11.29 points on the margin**, and **12.43 points on game totals**. A 20–20 prediction counts as an incorrect outright winner call. Latest predictions are used, including the original immutable Thursday ledger; postgame recalculations are excluded.

The largest correctable input defects were stale weather (Atlanta–Tampa Bay still used an October 27 forecast) and missed late-October trades. Tennessee–Cincinnati and Las Vegas–New York had much greater scoring than expected. Baltimore sustained two quarterback injuries during its game, which were not knowable outcomes before kickoff. The model still underrepresents some extreme game paths; no claim is made that these fixes explain all error.

## Changes in this release

- Refreshed 121 completed games, current team form, play-by-play efficiency, passing/rushing usage, pressure, special teams, coaching/penalty context and quarterback form through November 2. Existing independently evaluated player matchup features remain enabled. Older line/coverage/film observations retain their source dates.
- Reconciled all 32 rosters and seven completed in-season trades. Joe Fagnano is Baltimore’s current starter; Jackson and Huntley are excluded. Arizona uses Gardner Minshew. Reserve designations require activation evidence; returning to practice does not itself clear a player.
- Fixed injury performance leaking from an injured backup QB onto the selected starter. Fixed manual Available/healthy overrides still incurring injury penalties. Confirmed active players may still have conditional injury reductions.
- Injuries affect selected players, opportunity allocation, efficiency and early exits. Nelson to Seattle, Buckner to San Francisco and Simmons to Chicago now also affect blocking or defensive fronts after score calibration. These are small capped positional assumptions, reduced as post-trade games enter team form, **not fitted causal player values or proven accuracy gains**.
- Refreshed all 15 exact Week 9 venue/kickoff weather records November 4; indoor games use indoor conditions. These remain early forecasts.
- Week 8 displays each saved pregame prediction beside its actual final; postgame reruns never replace or grade the pregame record. Week 9 is the default scorecard. Archived forecasts and revisions are preserved.
- Restored score distributions and player averages in every new frozen forecast so the next audit can examine player volume and range coverage.

## Week 9 predictions

Each matchup uses 2,000 play simulations. Scores are rounded means, not the single most likely score. Win probabilities use the existing correction fitted on 2024–2025 forecasts; ties are preserved from simulation. Rounding can make percentages total 99.9% or 100.1%.

| Away | Predicted | Away win | Home | Predicted | Home win | Tie | QA |
|---|---:|---:|---|---:|---:|---:|---|
| JAX | 26 | 69.6% | BAL | 20 | 28.8% | 1.6% | consistent |
| CIN | 27 | 66.5% | ATL | 22 | 32.3% | 1.1% | consistent |
| DEN | 24 | 65.3% | CAR | 19 | 33.0% | 1.7% | consistent |
| DAL | 27 | 62.9% | IND | 23 | 35.9% | 1.2% | consistent |
| NYJ | 18 | 28.5% | KC | 25 | 70.1% | 1.4% | consistent |
| DET | 29 | 71.3% | MIA | 21 | 27.5% | 1.2% | consistent |
| CLE | 19 | 38.8% | NO | 23 | 59.8% | 1.4% | consistent |
| NYG | 22 | 52.2% | PHI | 21 | 46.0% | 1.8% | consistent |
| LA | 26 | 68.8% | WAS | 20 | 30.4% | 0.8% | consistent |
| HOU | 24 | 68.3% | LAC | 19 | 30.9% | 0.9% | consistent |
| LV | 17 | 17.8% | SF | 29 | 81.4% | 0.9% | review |
| GB | 22 | 45.3% | NE | 23 | 53.0% | 1.7% | review |
| ARI | 16 | 18.5% | SEA | 28 | 80.7% | 0.8% | consistent |
| TB | 20 | 43.2% | CHI | 23 | 55.7% | 1.1% | review |
| BUF | 24 | 62.6% | MIN | 20 | 36.1% | 1.2% | consistent |

QA means the baseline classifier and simulation were checked for disagreement and independent numerical score calibration. Consistent does not mean accurate; review flags remain visible. With 2,000 samples, raw win-rate sampling error can be about ±2.2 percentage points near 50%, in addition to much larger model/input uncertainty.

Flag details: LV–SF missed independent numerical calibration tolerance (home validation 29.12 versus 30.67 target, a 1.55-point gap). GB–NE and TB–CHI have baseline/simulation probability disagreements above 10 percentage points. These flags remain visible; none were forced into agreement.

## Week 8 predicted versus final

Scores are away–home. These are the actual latest eligible forecasts saved before each kickoff.

| Matchup | Predicted | Actual final |
|---|---:|---:|
| CAR at GB | 21–22 | 19–27 |
| BAL at BUF | 23–27 | 10–36 |
| TEN at CIN | 15–27 | 36–43 |
| ARI at DAL | 23–29 | 33–30 |
| MIN at DET | 22–24 | 15–19 |
| IND at JAX | 18–27 | 7–42 |
| LV at NYJ | 21–23 | 38–33 |
| CLE at PIT | 17–23 | 6–27 |
| ATL at TB | 21–23 | 0–16 |
| LAC at LA | 18–27 | 9–24 |
| KC at DEN | 20–20 | 24–28 |
| NE at MIA | 22–17 | 24–19 |
| PHI at WAS | 22–21 | 10–27 |
| CHI at SEA | 19–25 | 34–17 |

## Remaining uncertainty

Fourteen personnel reconciliation flags remain documented in ROSTER_INJURY_REVIEW.md. They include conflicting ownership feeds and reserve/activation evidence. Injury feeds can cap results at 25 entries per team. Current practice/body-part details were available for only some teams; missing details are not treated as healthy. Friday designations and game-day inactives are not yet available. Earlier film, coverage and timing metrics are dated.

This release does not refit new scoring weights to fourteen outcomes. Future research should evaluate severe-rain behavior, pressure with replacement QBs, finishing efficiency and correlated scoring tails on chronological holdouts across multiple seasons. Current fixes repair demonstrated defects and use fresh observations; only future evaluation can establish improved forecast accuracy.

## Verification and use

The deploy ZIP is self-contained for Cloudflare Pages, with index.html and _worker.js at the archive root. Deploy its extracted contents to replace the hosted build. This deliverable does not itself change a hosted site. The included worker refreshes supported data on access with hourly caching; no 10am scheduled job is installed. Saved pregame forecasts remain fixed.

Verification covers all 14 independent final-score agreements, historical archive preservation, 15 future forecasts, probability normalization, offline rendered scorecards, roster/quarterback/trade integrity, injury overrides, actual engine sensitivity, and current worker prediction parity. Detailed evidence, regression tests and source captures are included in the source/audit archive.

Read WEEK8_AUDIT.md for full diagnostics and source URLs; ROSTER_INJURY_REVIEW.md for every team and unresolved flag; PERSONNEL_MODEL_REVIEW.md for controlled sensitivity tests and limitations.