CoolFace
Datasetpublic

AdamRoch/nfl-play-normalization-v2

NFL Play Normalization V2 Data-Efficiency Curve This dataset contains four nested training sets with 3,125, 6,250, 12,500, and 25,000 unchanged real nflverse play descriptions from the 2019 through 2022 seasons. The v2 selection targets laterals, repeated fumbles, accepted penalties after turnovers, and descriptions with several yardage clauses. The curve manifest pins every dataset and point-manifest checksum. The review file explains the deterministic selection and the… See the full description on the dataset page: https://huggingface.co/datasets/AdamRoch/nfl-play-normalization-v2.

sourceHugging Facecc-by-4.0updated 29d agoView on Hugging Face
0likes40downloads
Dataset Card

NFL Play Normalization V2 Data-Efficiency Curve

This dataset contains four nested training sets with 3,125, 6,250, 12,500, and 25,000 unchanged real nflverse play descriptions from the 2019 through 2022 seasons. The v2 selection targets laterals, repeated fumbles, accepted penalties after turnovers, and descriptions with several yardage clauses.

The curve manifest pins every dataset and point-manifest checksum. The review file explains the deterministic selection and the fixed-data comparison with the corrected MVP sample.

Construction and revisions

  • —Source corpus SHA-256: 9129b63f8de1c45e864898cf4d6f0fcd453e59fc5c0587d05b59d54788a49957
  • —Canonical schema SHA-256: b5fc039929b7a93753f41a233579881f560ef83446f2aa0fe255fe4e9ddaa807
  • —Publication code revision: de8bba2ce956be1061e3f0d97e7b556abd9ed5ab
  • —Selection rule: multi-event-yardage-v1

The immutable Hugging Face dataset revision is recorded by scripts/publish_v2_dataset.py after upload because a repository cannot include the hash of the commit that contains that same hash.

Source attribution

nflverse data is licensed CC BY 4.0. Attribute nflverse and its upstream sources when using this dataset. The exact source releases are:

  • —2019: https://github.com/nflverse/nflverse-data/releases/download/pbp/play_by_play_2019.parquet with SHA-256 60c3067017db2d28a78f66a79b657268be8578d9a5288e6a827efdcd7fe42540
  • —2020: https://github.com/nflverse/nflverse-data/releases/download/pbp/play_by_play_2020.parquet with SHA-256 73b7dbf66fa8cb9356f58bf6b1f15a0fee197ecc10cf4983b640cb9679b15cb4
  • —2021: https://github.com/nflverse/nflverse-data/releases/download/pbp/play_by_play_2021.parquet with SHA-256 333ad34378e5339d5172717cc83378e908daf02c8699416ab3e17c2ec10f78d8
  • —2022: https://github.com/nflverse/nflverse-data/releases/download/pbp/play_by_play_2022.parquet with SHA-256 931121d8897779d7944e2a293e92ed8799c8e5cceef84096ac42339003fedc09

Intended use and limitations

Use this dataset to train NFL play-description normalizers against the included canonical JSON contract and to reproduce the four-point data-efficiency curve. It is not a replacement for nflverse structured fields. It does not cover spoken commentary, video, betting, fantasy, prediction, or coaching advice. The selection deliberately increases multi-event yardage examples, so it is not an unbiased sample of all NFL plays.