MLB Modeling Features (2023-2025)
41 pre-engineered features combining Pinnacle odds, advanced metrics, weather, pitcher stats, and park factors. Plug into any ML framework.
What's Inside
Every column, every row — here's exactly what you get.
Pitcher Intelligence
Starting pitcher xERA, throwing hand, batter vs SP wOBA matchups, xERA differential, and bullpen ERA (rolling).
Rolling Metrics
7-day rolling averages: pitch velocity, exit velocity, and HR rate. Pre-computed so you don't have to window yourself.
Park & Weather
Park run factors, temperature (°C), wind speed (kph), dome indicator. All features pre-joined — zero engineering needed.
Data Preview
5 sample rows| # | game_pk | date | home | away | ft_line | f5_line | spread | sp_home | xera_h | xera_a | bp_era_h | park_f | temp_c | dome |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 745444 | 2024-03-20 | SD | LAD | 9.0 | 4.5 | 1.5 | Yu Darvish | 3.42 | 3.18 | 3.85 | 0.97 | 21.3 | 0 |
| 2 | 746175 | 2024-03-21 | LAD | SD | 8.5 | 4.5 | -1.5 | Y. Yamamoto | 3.61 | 4.02 | 3.72 | 1.03 | 19.8 | 0 |
| 3 | 745039 | 2024-03-28 | TEX | CHC | 8.5 | 4.5 | -1.5 | N. Eovaldi | 3.89 | 3.55 | 4.12 | 1.08 | 24.1 | 1 |
| 4 | 745116 | 2024-03-28 | TB | TOR | 7.5 | 4.0 | -1.5 | Z. Eflin | 3.24 | 3.96 | 3.58 | 0.92 | 23.5 | 1 |
| 5 | 745283 | 2024-03-28 | SEA | BOS | 7.5 | 4.0 | -1.5 | L. Castillo | 3.15 | 4.21 | 3.41 | 0.95 | 11.2 | 1 |
Games Included
| Season | Games | of 2,430 |
|---|---|---|
| 2023 | 1,822 | 75.0% |
| 2024 | 2,296 | 94.5% |
| 2025 | 2,105 | 86.6% |
| Total | 6,223 | 91.7% |
*Each MLB season has 2,430 scheduled games. Games with incomplete odds or missing scores are excluded — that's why every included row has zero missing values.
Schema (Key Columns)
| Column | Description |
|---|---|
| game_pk / game_date | Game ID + date |
| ft_line / f5_line / spread | Pinnacle open + current |
| sp_home_name / sp_away_name | Starting pitcher names |
| sp_home_xera / sp_away_xera | Expected ERA (xERA) |
| xera_diff | xERA differential |
| bullpen_era_home / _away | Bullpen ERA (rolling) |
| velo_* / exit_velo_* | 7-day rolling pitcher metrics |
| woba_home_vs_sp | Lineup wOBA vs opposing SP |
| park_factor | Park run factor |
| temp_c / wind_kph / dome | Weather + dome indicator |
| score_home / total_runs | Target variables |
Save 40+ hours of data engineering. Building this dataset from scratch means writing API integrations, handling rate limits, cleaning edge cases, filling gaps, and joining multiple sources. We've done all of that. You get the clean result — ready to load into pandas in one line.
Built For
Data Scientists
Train predictive models with 41 pre-engineered features. Works with scikit-learn, XGBoost, LightGBM, CatBoost, PyTorch, R.
Skip Feature Engineering
All features pre-computed and pre-joined. No API integrations, no data cleaning, no missing values to fill.
Students & Researchers
Study sports analytics with real data. Perfect for theses, projects, and Kaggle competitions.
Included Files
Each dataset in CSV and Parquet — per-season splits + combined bundle.
Frequently Asked Questions
What file formats are included?
How do I download the files?
What does "Zero NULLs" mean?
Where do the odds come from?
Can I use this data commercially?
Do I still need to do feature engineering?
What ML frameworks are compatible?
Ready to build?
One file. One purchase. Infinite backtests.
