SebastianAndreu commited on
Commit
fc91e33
·
verified ·
1 Parent(s): 72fc7cd

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +219 -0
README.md CHANGED
@@ -208,4 +208,223 @@ configs:
208
  data_files:
209
  - split: full
210
  path: data/full-*
 
 
 
211
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
208
  data_files:
209
  - split: full
210
  path: data/full-*
211
+ license: mit
212
+ language:
213
+ - en
214
  ---
215
+
216
+ ---
217
+ license: mit
218
+ task_categories:
219
+ - tabular-regression
220
+ - tabular-classification
221
+ tags:
222
+ - nfl
223
+ - football
224
+ - sports-analytics
225
+ - wide-receiver
226
+ - play-by-play
227
+ size_categories:
228
+ - 1K<n<10K
229
+ ---
230
+
231
+ # NFL Wide Receiver Performance Dataset (2015)
232
+
233
+ ## Dataset Description
234
+
235
+ This dataset contains comprehensive wide receiver performance statistics derived from NFL play-by-play data for the **2015**. It includes game-level metrics, situational targeting patterns, defensive adjustments, and advanced efficiency calculations.
236
+
237
+ ### Dataset Summary
238
+
239
+ - **Season**: 2015
240
+ - **Records**: ~4.31k player-game observations
241
+ - **Features**: 100+ columns including:
242
+ - Core statistics (targets, receptions, yards, touchdowns)
243
+ - Quarter-by-quarter breakdowns
244
+ - Win probability bucketed performance
245
+ - Defensive strength adjustments
246
+ - Situational metrics (red zone, high leverage, down-and-distance)
247
+ - Team share metrics (target share, air yards share, WOPR)
248
+ - Efficiency metrics (aDOT, yards per target, catch rate)
249
+ - Weather and venue conditions
250
+
251
+ ### Supported Tasks
252
+
253
+ - **Receiving Yards Prediction**: Predict receiving yards for upcoming games
254
+ - **Target Share Analysis**: Model player opportunity distribution
255
+ - **Performance Forecasting**: Project future player performance
256
+ - **Matchup Analysis**: Evaluate player-defense matchups
257
+
258
+ ## Dataset Structure
259
+
260
+ ### Data Fields
261
+
262
+ **Key Identifiers:**
263
+ - `game_id`: Unique game identifier
264
+ - `receiver_player_id`: NFL GSIS player ID
265
+ - `receiver_player_name`: Player display name
266
+ - `passer_player_id`: Quarterback player ID
267
+ - `season`: NFL season year
268
+ - `week`: Week number
269
+
270
+ **Core Statistics:**
271
+ - `targets`: Total pass attempts targeting the receiver
272
+ - `receptions`: Completed receptions
273
+ - `receiving_yards`: Total receiving yards
274
+ - `tds`: Receiving touchdowns
275
+ - `air_yards`: Total air yards on targets
276
+ - `yac`: Yards after catch
277
+
278
+ **Quarter Breakdowns:**
279
+ - `yards_Q1`, `yards_Q2`, `yards_Q3`, `yards_Q4`: Yards by quarter
280
+ - `receptions_Q1-4`: Receptions by quarter
281
+ - `targets_Q1-4`: Targets by quarter
282
+
283
+ **Win Probability Buckets:**
284
+ - `yards_wp_<25`, `yards_wp_25_45`, etc.: Performance in different game situations
285
+ - Similar breakdowns for receptions and targets
286
+
287
+ **Share Metrics:**
288
+ - `target_share`: Player's share of team targets
289
+ - `air_yards_share`: Player's share of team air yards
290
+ - `yard_share`: Player's share of team receiving yards
291
+ - `reception_share`: Player's share of team receptions
292
+ - `wopr`: Weighted Opportunity Rating (0.7 × target_share + 0.3 × air_yards_share)
293
+
294
+ **Efficiency Metrics:**
295
+ - `aDOT`: Average depth of target
296
+ - `yards_per_target`: Receiving yards per target
297
+ - `catch_rate`: Reception percentage
298
+ - `yac_per_rec`: Yards after catch per reception
299
+ - `explosive_rec_pct`: Percentage of receptions ≥15 yards
300
+ - `first_down_pct`: Percentage of receptions resulting in first downs
301
+
302
+ **Defensive Adjustments:**
303
+ - `def_targets_dev`: Defense targets allowed vs league average
304
+ - `def_receptions_dev`: Defense receptions allowed vs league average
305
+ - `def_yards_dev`: Defense yards allowed vs league average
306
+ - `def_tds_dev`: Defense TDs allowed vs league average
307
+ - `def_epa_dev`: Defense EPA allowed vs league average
308
+ - `adj_epa`: Defense-adjusted Expected Points Added
309
+ - `adj_epa_per_target`: Defense-adjusted EPA per target
310
+
311
+ **Situational Metrics:**
312
+ - `red_zone_targets`: Targets inside the 20-yard line
313
+ - `end_zone_targets`: Targets in the end zone
314
+ - `third_down_targets`: Targets on 3rd down
315
+ - `fourth_down_targets`: Targets on 4th down
316
+ - `high_leverage_targets`: Targets in high-leverage situations (WP < 0.25 or > 0.75)
317
+ - `red_zone_share`: Player's share of team red zone targets
318
+ - `third_down_share`: Player's share of team 3rd down targets
319
+
320
+ **Game Context:**
321
+ - `posteam`: Player's team (encoded 1-32)
322
+ - `defteam`: Opposing defense (encoded 1-32)
323
+ - `home_team`: Home team (encoded 1-32)
324
+ - `away_team`: Away team (encoded 1-32)
325
+ - `home_flag`: 0 if home, 1 if away
326
+ - `pregame_spread`: Betting line point spread
327
+ - `pregame_total`: Betting line total points
328
+ - `avg_score_diff`: Average score differential when targeted
329
+ - `avg_quarter`: Average quarter when targeted
330
+ - `trailing_pct`: Percentage of targets while trailing
331
+ - `leading_pct`: Percentage of targets while leading
332
+
333
+ **Weather & Venue:**
334
+ - `surface`: Playing surface type (encoded 0-6)
335
+ - `is_dome`: 1 if indoor, 0 if outdoor
336
+ - `is_rain`: 1 if rainy conditions
337
+ - `is_snow`: 1 if snowy conditions
338
+ - `is_clear`: 1 if clear conditions
339
+ - `temp_f`: Temperature in Fahrenheit
340
+ - `humidity_pct`: Humidity percentage
341
+ - `wind_mph`: Wind speed in miles per hour
342
+
343
+ **QB Context:**
344
+ - `qb_completions`: Quarterback's completions that game
345
+ - `qb_attempts`: Quarterback's attempts that game
346
+ - `qb_comp_pct`: Quarterback's completion percentage
347
+ - `qb_air_yards`: Quarterback's average air yards
348
+ - `qb_cpoe`: Quarterback's completion percentage over expected
349
+
350
+ **Advanced Metrics:**
351
+ - `epa`: Expected Points Added
352
+ - `wpa`: Win Probability Added
353
+ - `success_rate`: Percentage of successful plays (EPA > 0 or YPT > 0.5)
354
+ - `big_play_rate`: Percentage of plays ≥20 yards
355
+ - `explosive_plays`: Count of plays ≥20 yards
356
+ - `first_downs`: First downs generated
357
+ - `consistency_score`: mean_adj_epa / std_adj_epa
358
+ - `inverse_volatility`: 1 / std_adj_epa
359
+ - `season_adj_epa_per_target`: Season-level defense-adjusted EPA per target
360
+ - `wp_var`: Variance in win probability across targets
361
+ - `target_share_std`: Standard deviation of target share across games
362
+
363
+ ### Data Splits
364
+
365
+ This dataset does not include pre-defined splits. Users should create their own train/validation/test splits based on their use case:
366
+ - **Time-based split**: Use early weeks for training, later weeks for validation/testing
367
+ - **Cross-validation**: K-fold cross-validation across games
368
+ - **Season holdout**: Train on this season, test on future seasons
369
+
370
+ ### Dataset Creation
371
+
372
+ #### Source Data
373
+
374
+ 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.
375
+
376
+ #### Data Processing
377
+
378
+ The dataset was created through two complementary processing pipelines:
379
+
380
+ 1. **Pipeline A (Defensive Adjustments)**:
381
+ - Calculates defense-adjusted performance metrics
382
+ - Adds situational targeting patterns
383
+ - Includes QB context and team-level statistics
384
+ - Incorporates weather and venue conditions
385
+
386
+ 2. **Pipeline B (Temporal & Situational)**:
387
+ - Generates quarter-by-quarter breakdowns
388
+ - Creates win probability bucketed statistics
389
+ - Computes team share metrics and WOPR
390
+ - Calculates season-level consistency metrics
391
+
392
+ Both pipelines were merged to create a comprehensive feature set.
393
+
394
+ ## Considerations for Using the Data
395
+
396
+ ### Social Impact
397
+
398
+ This dataset is intended for:
399
+ - Sports analytics and research
400
+ - Fantasy football decision-making
401
+ - Educational purposes in machine learning and sports statistics
402
+
403
+ **Not intended for:**
404
+ - Real-money gambling (use responsibly)
405
+ - Player evaluation for contract negotiations
406
+ - Any decision-making that could impact player careers
407
+
408
+ ### Discussion of Biases
409
+
410
+ - **Opportunity bias**: Statistics heavily dependent on team offensive scheme and QB quality
411
+ - **Injury data**: Dataset does not account for injuries that may affect performance
412
+ - **Sample size**: Players with limited playing time have less reliable statistics
413
+ - **Game script**: Performance metrics influenced by whether team is winning/losing
414
+ - **Strength of schedule**: Not all defensive matchups are equal, though some adjustment is included
415
+
416
+ ### Limitations
417
+
418
+ - **Historical data only**: Does not predict future performance definitively
419
+ - **Missing context**: Does not include play design, route running, or other qualitative factors
420
+ - **Weather parsing**: Temperature/wind/humidity may be missing or inaccurate for some games
421
+ - **Roster changes**: Does not account for mid-season team changes or trades
422
+ - **Playoff games**: May or may not include playoff data depending on the year
423
+
424
+ ## Additional Information
425
+
426
+ ### Acknowledgments
427
+
428
+ - **Data Source**: [nflverse/nflverse-data](https://github.com/nflverse/nflverse-data/releases/tag/pbp)
429
+ - **AI Assistance**: Code development assisted by Claude (Anthropic)
430
+ - **Course**: CMU 24-679: Designing and Deploying AI/ML Systems