File size: 3,059 Bytes
9cb3b03 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 | ---
license: cc-by-nc-sa-4.0
pretty_name: OpenCommand
language:
- en
tags:
- baseball
- sports-analytics
- computer-vision
- statcast
size_categories:
- 1M<n<10M
configs:
- config_name: command_scores_2025
data_files: 2025/command_scores.csv
- config_name: command_scores_2026
data_files: 2026/command_scores.csv
- config_name: targets_2025
data_files: 2025/targets.csv.gz
- config_name: targets_2026
data_files: 2026/targets.csv.gz
---
# OpenCommand
Catcher targets and pitcher command for MLB, inferred from broadcast video.
OpenCommand scores **command** using the pitch location's distance from target.
This dataset contains the 2025/2026 computer vision object detections, every intermediate the pipeline writes, and the resulting **command scores**. The pipeline itself and the full method write-up live at [github.com/tomdoyo/open-command](https://github.com/tomdoyo/open-command).
## Download
```
hf download tomdoyo/open-command --repo-type dataset --local-dir data
```
That puts the tree where the pipeline expects it, so any stage re-runs from the code repo:
```
python src/solve_glove_locations.py 2026
```
One file instead of the whole tree:
```
hf download tomdoyo/open-command 2026/command_scores.csv --repo-type dataset
```
## Data
Each season lives under `<year>/`.
**Keys:** `(game_pk, play_id)` identify a pitch.
Raw detections (in `<year>/raw/`) are produced using YOLO11 glove/ball/strikezone detector models, with postprocessing based on detection confidence.
| File (per season) | One row per | Contents |
|---|---|---|
| `pbp_info.csv.gz` | pitch | Statcast 9-parameter trajectory, `sz_top`/`sz_bot`, plate location, pitcher, pitch type, and the game_date/type/venue |
| `raw/gloveball_tracks/<game_pk>.csv.gz` | frame | glove + ball detections (pixels on screen) |
| `raw/strikezone_tracking.csv.gz` | clip | broadcast strikezone detections (pixels on screen) |
| `camera_poses.csv.gz` | clip | camera pose (+ vote diagnostics & reprojection accuracies) |
| `glove_locations/<game_pk>.csv.gz` | detection | solved glove location (real-world) |
| `targets.csv.gz` | clip | naive/inferred targets |
| `command_scores.csv` | pitcher, pitch type | n, naive and inferred median miss |
`fg_pitching_2025.csv.gz` and `fg_pitching_2026.csv.gz` sit at the root, carrying the FanGraphs season lines command is validated against.
## License & citation
Everything here is released under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/): use it, build on it, publish with it, with **attribution** (cite OpenCommand) and **not commercially**.
Data derived from MLB broadcast video and Statcast public feeds. MLB and Statcast are trademarks of MLB Advanced Media, L.P.; this project is not affiliated with MLB.
```bibtex
@dataset{kim_opencommand,
author = {Kim, Tom},
title = {OpenCommand: open catcher-target and pitcher-command data for MLB},
year = {2026},
url = {https://open-command.com},
note = {https://github.com/tomdoyo/open-command}
}
```
|