PICK'EM
Methodology

A model that has to keep receipts.

Pick'em separates the quantitative model from Bryan's final decision. That makes it possible to measure whether human judgment adds value or just adds Cowboys bias.

Production model · Week 5+

Current Season EPA v2.0.0

  • • 65% season-to-date net EPA/play from the current 2026 season
  • • 25% net EPA/play across each team's most recent two games
  • • 10% turnover margin per game
  • • Small fixed home-field adjustment

For each target week, v2 uses only 2026 results available through the prior week. Week 5 therefore uses Weeks 1–4. Each prediction stores its model version, source context, feature snapshot, rationale, and generation timestamp.

What v2 does not claim

  • • No injury or player-availability adjustment yet.
  • • No sportsbook or betting-market input.
  • • No hidden AI judgment or manual team boosts.
  • • Model Edge is a 0–100 separation index, not a calibrated real-world win probability.
Preserved history

EPA Baseline v1.0.0 · Weeks 1–4

V1 used the complete 2025 season because the 2026 sample was initially too small. Those predictions remain immutable historical receipts. Model v2 begins with Week 5; switching versions never rewrites earlier picks.

Accountability rules

  1. 1. Bryan picks publish before kickoff.
  2. 2. Published Bryan picks receive timestamps and cannot be edited.
  3. 3. Model predictions are versioned and stored with their feature snapshots.
  4. 4. Existing predictions are never overwritten by a later model version.
  5. 5. Season scoring uses the model version that was active for that week.
  6. 6. Every completed result counts.
  7. 7. Third-party benchmarks are sourced, not silently copied.
National Expert Benchmark

The bar is leading national NFL experts.

The public benchmark uses the strongest published straight-up season record in the CBS Sports national expert panel. It is intentionally a high bar rather than a league-average or random-pick baseline. The source is identified publicly so the comparison can be checked independently.

View CBS Sports expert picks ↗
Cowboys Bias Index

Yes, this is a real metric.

The Bias Index is the percentage of published Cowboys games where the model version active for that week picked against Dallas and Bryan still chose Dallas. Homer Tax is the net number of wins Bryan lost versus that active model on completed Cowboys overrides. Both are calculated from stored picks, versioned model predictions, and final game results.