Philadelphia Eagles at Tennessee Titans

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: punt; expected WP go/FG/punt: 0.18490/ NA/0.22157
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Chosen action ambiguous

0 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: go; preferred baseline action: go; expected WP go/FG/punt: 0.34167/0.27176/ NA
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: punt; expected WP go/FG/punt: 0.24654/ NA/0.28200
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: go; preferred baseline action: go; expected WP go/FG/punt: 0.72764/0.67939/0.69783
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: fg; preferred baseline action: fg; expected WP go/FG/punt: 0.11925/0.16902/ NA
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: punt; expected WP go/FG/punt: 0.71756/ NA/0.77066
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: go; expected WP go/FG/punt: 0.17062/ NA/0.16820
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Kickoff assumption sensitive

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: fg; preferred baseline action: punt; expected WP go/FG/punt: 0.78981/0.78775/0.79189
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: punt; expected WP go/FG/punt: 0.58129/ NA/0.60918
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: punt; preferred baseline action: punt; expected WP go/FG/punt: 0.32778/ NA/0.36759
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

Close call

1 of 1 eligible opportunities modeled.

• Spread-adjusted nfl4th blended WP; distinct from officiating WP.
• Six seconds per action; no added successful-conversion runoff.
• No go-for-it turnover returns; kicking ignores kicker identity/weather; ordinary kickoff continuations only.
• Close-call tolerance is a product rule of 1 percentage point, not model uncertainty.
• Upstream training cutoff unverified; current-season validation status is documented separately.
• Kickoff continuation scenarios: standard_own35, touchback_own20, prior_return_start_35, prior_return_start_31, prior_return_start_26; sensitivity is not a confidence interval.
• Chosen action: fg; preferred baseline action: fg; expected WP go/FG/punt: 0.46332/0.59305/ NA
• A supported value is a model-relative estimate in a tested domain, not a proven causal cost.

Model: under-review-nfl4th-2026-v1

0 of 0 eligible opportunities modeled.

• Ordinary single in-bounds fumbles only; recovery creation separated from recovery outcome.
• Recovery residual only; no unsupported recovery WP branches.
• Includes both own and opposing fumbles; two team views share the same events and sum to zero.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-5d13bdb7c25f-team-recovery-v1

0 of 0 eligible opportunities modeled.

• Ordinary single in-bounds fumbles only; recovery creation separated from recovery outcome.
• Recovery residual only; no unsupported recovery WP branches.
• Includes both own and opposing fumbles; two team views share the same events and sum to zero.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-5d13bdb7c25f-team-recovery-v1

2 of 2 eligible opportunities modeled.

• Nullified/blocked attempts and returns excluded; blocked kicks are not automatically kicker errors.
• Distance, era and reported roof; no weather or kicker-history adjustment in this model.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-fc871868cb91

1 of 1 eligible opportunities modeled.

• Nullified/blocked attempts and returns excluded; blocked kicks are not automatically kicker errors.
• Distance, era and reported roof; no weather or kicker-history adjustment in this model.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-fc871868cb91

2 of 2 eligible opportunities modeled.

• Nullified/blocked attempts and returns excluded; blocked kicks are not automatically kicker errors.
• Distance, era and reported roof; no weather or kicker-history adjustment in this model.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-1b26bba80d09

3 of 3 eligible opportunities modeled.

• Nullified/blocked attempts and returns excluded; blocked kicks are not automatically kicker errors.
• Distance, era and reported roof; no weather or kicker-history adjustment in this model.
• Chronological training: 2015,2016,2017,2018,2019,2020,2021,2022; calibration: 2023

Model: under-review-contextual-baselines-v1-1b26bba80d09

20 of 20 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

37 of 37 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

20 of 20 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

37 of 37 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

2 of 2 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

3 of 3 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

20 of 20 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

37 of 37 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

20 of 20 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

37 of 37 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

20 of 20 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

37 of 37 eligible opportunities modeled.

• FTN charting labels, not calibrated outcome probabilities or error-attributable WPA.
• Coverage counts nonmissing charted labels; uncharted plays are not zero errors.
• FTN attribution and CC-BY-SA obligations apply.

Model: FTN-charting

54 of 54 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 1 across 54 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

54 of 54 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 1 across 54 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

40 of 40 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 0 across 40 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

40 of 40 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 0 across 40 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

76 of 76 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 0 across 76 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

76 of 76 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 3 across 76 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

23 of 23 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 0 across 23 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1

23 of 23 eligible opportunities modeled.

• Called/enforced penalties, not adjudicated foul correctness.
• Family-specific opportunity counts; prior-game team/opponent tendencies; crew effects not normalized away.
• Observed 0 across 23 opportunities; penalty-generated first downs 0.

Model: under-review-contextual-baselines-v1