--- dataset_info: features: - name: game_id dtype: string - name: receiver_player_id dtype: string - name: receiver_player_name dtype: string - name: passer_player_id dtype: string - name: defteam dtype: int64 - name: targets dtype: float64 - name: receptions dtype: float64 - name: receiving_yards dtype: float64 - name: air_yards dtype: float64 - name: yac dtype: float64 - name: tds dtype: float64 - name: epa dtype: float64 - name: wpa dtype: float64 - name: avg_depth dtype: float64 - name: catch_rate dtype: float64 - name: posteam dtype: int64 - name: team_pass_attempts dtype: float64 - name: team_air_yards dtype: float64 - name: team_epa dtype: float64 - name: air_yard_share dtype: float64 - name: target_share dtype: float64 - name: yards_per_target dtype: float64 - name: def_targets_dev dtype: float64 - name: def_receptions_dev dtype: float64 - name: def_yards_dev dtype: float64 - name: def_tds_dev dtype: float64 - name: def_epa_dev dtype: float64 - name: red_zone_targets dtype: int64 - name: end_zone_targets dtype: int64 - name: third_down_targets dtype: int64 - name: fourth_down_targets dtype: int64 - name: high_leverage_targets dtype: int64 - name: qb_completions dtype: float64 - name: qb_attempts dtype: float64 - name: qb_air_yards dtype: float64 - name: qb_cpoe dtype: float64 - name: qb_comp_pct dtype: float64 - name: avg_score_diff dtype: float64 - name: avg_quarter dtype: float64 - name: adot dtype: float64 - name: yac_per_reception dtype: float64 - name: td_rate dtype: float64 - name: explosive_plays dtype: float64 - name: first_downs dtype: float64 - name: trailing_pct dtype: float64 - name: leading_pct dtype: float64 - name: wp_var dtype: float64 - name: target_share_std dtype: float64 - name: surface dtype: int64 - name: is_dome dtype: int64 - name: is_rain dtype: int64 - name: is_snow dtype: int64 - name: is_clear dtype: int64 - name: temp_f dtype: float64 - name: humidity_pct dtype: float64 - name: wind_mph dtype: float64 - name: success_rate dtype: float64 - name: big_play_rate dtype: float64 - name: reception_std dtype: float64 - name: second_and_long_targets dtype: float64 - name: third_and_medium_targets dtype: float64 - name: avg_start_yardline dtype: float64 - name: avg_target_depth_vs_qb dtype: float64 - name: yards_Q1 dtype: float64 - name: yards_Q2 dtype: float64 - name: yards_Q3 dtype: float64 - name: yards_Q4 dtype: float64 - name: receptions_Q1 dtype: float64 - name: receptions_Q2 dtype: float64 - name: receptions_Q3 dtype: float64 - name: receptions_Q4 dtype: float64 - name: targets_Q1 dtype: float64 - name: targets_Q2 dtype: float64 - name: targets_Q3 dtype: float64 - name: targets_Q4 dtype: float64 - name: lost_yards_due_to_penalty dtype: float64 - name: yards_wp_<25 dtype: float64 - name: yards_wp_25_45 dtype: float64 - name: yards_wp_45_55 dtype: float64 - name: yards_wp_55_75 dtype: float64 - name: yards_wp_>75 dtype: float64 - name: yards_wp_NA dtype: int64 - name: receptions_wp_<25 dtype: float64 - name: receptions_wp_25_45 dtype: float64 - name: receptions_wp_45_55 dtype: float64 - name: receptions_wp_55_75 dtype: float64 - name: receptions_wp_>75 dtype: float64 - name: receptions_wp_NA dtype: int64 - name: targets_wp_<25 dtype: int64 - name: targets_wp_25_45 dtype: int64 - name: targets_wp_45_55 dtype: int64 - name: targets_wp_55_75 dtype: int64 - name: targets_wp_>75 dtype: int64 - name: targets_wp_NA dtype: int64 - name: home_team dtype: int64 - name: away_team dtype: int64 - name: pregame_spread dtype: float64 - name: pregame_total dtype: float64 splits: - name: full num_bytes: 3498382 num_examples: 4314 download_size: 661109 dataset_size: 3498382 configs: - config_name: default data_files: - split: full path: data/full-* license: mit language: - en --- --- license: mit task_categories: - tabular-regression - tabular-classification tags: - nfl - football - sports-analytics - wide-receiver - play-by-play size_categories: - 1K 0.75) - `red_zone_share`: Player's share of team red zone targets - `third_down_share`: Player's share of team 3rd down targets **Game Context:** - `posteam`: Player's team (encoded 1-32) - `defteam`: Opposing defense (encoded 1-32) - `home_team`: Home team (encoded 1-32) - `away_team`: Away team (encoded 1-32) - `home_flag`: 0 if home, 1 if away - `pregame_spread`: Betting line point spread - `pregame_total`: Betting line total points - `avg_score_diff`: Average score differential when targeted - `avg_quarter`: Average quarter when targeted - `trailing_pct`: Percentage of targets while trailing - `leading_pct`: Percentage of targets while leading **Weather & Venue:** - `surface`: Playing surface type (encoded 0-6) - `is_dome`: 1 if indoor, 0 if outdoor - `is_rain`: 1 if rainy conditions - `is_snow`: 1 if snowy conditions - `is_clear`: 1 if clear conditions - `temp_f`: Temperature in Fahrenheit - `humidity_pct`: Humidity percentage - `wind_mph`: Wind speed in miles per hour **QB Context:** - `qb_completions`: Quarterback's completions that game - `qb_attempts`: Quarterback's attempts that game - `qb_comp_pct`: Quarterback's completion percentage - `qb_air_yards`: Quarterback's average air yards - `qb_cpoe`: Quarterback's completion percentage over expected **Advanced Metrics:** - `epa`: Expected Points Added - `wpa`: Win Probability Added - `success_rate`: Percentage of successful plays (EPA > 0 or YPT > 0.5) - `big_play_rate`: Percentage of plays ≥20 yards - `explosive_plays`: Count of plays ≥20 yards - `first_downs`: First downs generated - `consistency_score`: mean_adj_epa / std_adj_epa - `inverse_volatility`: 1 / std_adj_epa - `season_adj_epa_per_target`: Season-level defense-adjusted EPA per target - `wp_var`: Variance in win probability across targets - `target_share_std`: Standard deviation of target share across games ### Data Splits This dataset does not include pre-defined splits. Users should create their own train/validation/test splits based on their use case: - **Time-based split**: Use early weeks for training, later weeks for validation/testing - **Cross-validation**: K-fold cross-validation across games - **Season holdout**: Train on this season, test on future seasons ### Dataset Creation #### Source Data Raw play-by-play data sourced from [nflverse](https://github.com/nflverse/nflverse-data/releases/tag/pbp), which aggregates official NFL data with additional features. #### Data Processing The dataset was created through two complementary processing pipelines: 1. **Pipeline A (Defensive Adjustments)**: - Calculates defense-adjusted performance metrics - Adds situational targeting patterns - Includes QB context and team-level statistics - Incorporates weather and venue conditions 2. **Pipeline B (Temporal & Situational)**: - Generates quarter-by-quarter breakdowns - Creates win probability bucketed statistics - Computes team share metrics and WOPR - Calculates season-level consistency metrics Both pipelines were merged to create a comprehensive feature set. ## Considerations for Using the Data ### Social Impact This dataset is intended for: - Sports analytics and research - Fantasy football decision-making - Educational purposes in machine learning and sports statistics **Not intended for:** - Real-money gambling (use responsibly) - Player evaluation for contract negotiations - Any decision-making that could impact player careers ### Discussion of Biases - **Opportunity bias**: Statistics heavily dependent on team offensive scheme and QB quality - **Injury data**: Dataset does not account for injuries that may affect performance - **Sample size**: Players with limited playing time have less reliable statistics - **Game script**: Performance metrics influenced by whether team is winning/losing - **Strength of schedule**: Not all defensive matchups are equal, though some adjustment is included ### Limitations - **Historical data only**: Does not predict future performance definitively - **Missing context**: Does not include play design, route running, or other qualitative factors - **Weather parsing**: Temperature/wind/humidity may be missing or inaccurate for some games - **Roster changes**: Does not account for mid-season team changes or trades - **Playoff games**: May or may not include playoff data depending on the year ## Additional Information ### Acknowledgments - **Data Source**: [nflverse/nflverse-data](https://github.com/nflverse/nflverse-data/releases/tag/pbp) - **AI Assistance**: Code development assisted by Claude (Anthropic) - **Course**: CMU 24-679: Designing and Deploying AI/ML Systems