README.md

---
license: other
task_categories:
  - tabular-classification
language:
  - en
tags:
  - cricket
  - cricsheet
  - parquet
  - aicrix
size_categories:
  - 1M<n<10M
---

# AICrix — processed cricket data (Parquet)

Processed cricket tables in **Apache Parquet** (ZSTD compression), using the **AICrix relational schema** (`db/migrations/001_cricket_schema.sql`).

## Files

| File | Description |
|------|-------------|
| `continents.parquet` | Continents |
| `countries.parquet` | Countries |
| `leagues.parquet` | Competitions |
| `seasons.parquet` | Seasons |
| `stages.parquet` | Tournament stages |
| `venues.parquet` | Venues |
| `positions.parquet` | Player positions |
| `teams.parquet` | Teams |
| `players.parquet` | Players |
| `squad.parquet` | Season squads |
| `officials.parquet` | Umpires / referees |
| `fixtures.parquet` | Matches |
| `fixture_dl_data.parquet` | D/L method rows |
| `runs.parquet` | Innings team scores |
| `batting_scoreboards.parquet` | Batting cards |
| `bowling_scoreboards.parquet` | Bowling cards |
| `balls.parquet` | Ball-by-ball (largest table) |
| `lineups.parquet` | Match lineups |
| `standings.parquet` | Table standings |
| `team_rankings.parquet` | ICC-style rankings |

## Usage

```python
import pandas as pd

fixtures = pd.read_parquet("fixtures.parquet")
print(fixtures[["id", "type", "starting_at"]].head())
```

With DuckDB (SQL over parquet):

```python
import duckdb

duckdb.sql("SELECT id, type FROM 'fixtures.parquet' LIMIT 5").show()
```

Download from the Hub:

```bash
hf download veloria-tech/aicrix --repo-type dataset --local-dir ./aicrix
```

## Schema notes

- Fixture / player / official IDs are **strings** (Cricsheet registry compatibility).
- Innings: `S1`–`S4`; ball uniqueness: `(fixture_id, inning, over_number, ball_number)`.
- In queries: `localteam_id` = home, `visitorteam_id` = away.

## Source

Derived from Cricsheet JSON (`data/raw` → NDJSON pipeline → Parquet export).