Dataset Viewer
Auto-converted to Parquet Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
diagnostic_type: large_string
target_nominal: double
empirical_rate: double
error: double
details: large_string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 879
to
{'discipline': Value('large_string'), 'model': Value('large_string'), 'n_samples': Value('int64'), 'mae': Value('float64'), 'rmse': Value('float64'), 'pearson_r': Value('float64'), 'spearman_rho': Value('float64'), 'kendall_tau': Value('float64'), 'winkler_90': Value('float64'), 'crps': Value('float64')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/parquet/parquet.py", line 220, in _generate_tables
                  yield Key(file_idx, batch_idx), self._cast_table(pa_table)
                                                  ~~~~~~~~~~~~~~~~^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/parquet/parquet.py", line 156, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              diagnostic_type: large_string
              target_nominal: double
              empirical_rate: double
              error: double
              details: large_string
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 879
              to
              {'discipline': Value('large_string'), 'model': Value('large_string'), 'n_samples': Value('int64'), 'mae': Value('float64'), 'rmse': Value('float64'), 'pearson_r': Value('float64'), 'spearman_rho': Value('float64'), 'kendall_tau': Value('float64'), 'winkler_90': Value('float64'), 'crps': Value('float64')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Dataset Card for CricEngine AI Dataset

1. Dataset Summary

The CricEngine AI Dataset is a comprehensive, open-access, point-in-time cricket analytics and predictive data lake. It provides normalized ball-by-ball match data, rolling multi-window feature stores, Bayesian latent skill ratings, probabilistic forecast distributions, and Value Over Replacement Player (VORP) metrics across all major men's professional cricket competitions.

  • Primary Source: Cricsheet (https://cricsheet.org)
  • Time Coverage: 2002 – 2026
  • Match Formats: Test Matches, One-Day Internationals (ODIs), T20 Internationals (T20Is), Indian Premier League (IPL), and Syed Mushtaq Ali Trophy (SMAT).
  • Match Count: 9,085 matches (4.38+ million deliveries).
  • Release Version: v1.0.0 (Immutable approved data state).

2. Dataset Structure & Components

The dataset is partitioned into four decoupled, versioned tiers:

A. Canonical DuckDB Relational Database (cricengine.duckdb)

  • Format: DuckDB Columnar Relational File (.duckdb v2.0.0)
  • Size: 383 MB (401,616,896 bytes)
  • SHA-256: a6a99617719a93578169bac670f75e394d9c2c84f7cffa3ab233fc97c3067aae
  • Total Tables: 14 relational tables in Third Normal Form (3NF).
  • Total Rows: 4,907,393 rows.
  • Key Tables: dim_match, dim_match_context, dim_player_master, fact_innings, fact_ball_by_ball, fact_match_participation, fact_delivery_dismissal, fact_dismissal_fielder.

B. Feature Store (feature_store_v3b)

  • Format: Apache Parquet (Snappy compressed, Hive partitioned).
  • Files: 1,164 files (~117 MB).
  • Total Records: 1,605,672 rows.
  • Subsets:
    • player_match: 200,709 rows (51 columns) of match-level performance metrics.
    • rolling_match: 602,127 rows (34 columns) across 5, 10, and 20-match moving windows with exponential decay and empirical shrinkage.
    • rolling_time: 602,127 rows (19 columns) across 30, 90, and 365 calendar-day windows.
    • contextual_features: 200,709 rows (18 columns) of venue park factors and opponent quality indices.
    • contextual_baselines: Reference tables for park factors and phase baselines.

C. Player Rating Store (rating_store_v4)

  • Format: Apache Parquet (Hive partitioned by format).
  • Files: 145 files (12.5 MB).
  • Model: Context-Aware Hierarchical Bayesian State-Space (CA-HBSS).
  • Subsets:
    • pre_match_snapshots: 19,642 pre-match snapshots (D-1 guarantee strictly frozen prior to match start).
    • player_rating_current: 9,406 active format states across batting, bowling, fielding, and all-round composite roles.
    • player_rating_history: 73,039 longitudinal rating progression checkpoints.

D. Predictive Store (predictive_store_v4c)

  • Format: Apache Parquet (Hive partitioned by format and year).
  • Files: 86 files (2.8 MB).
  • Model: Two-Stage Hurdle / Zero-Inflated Poisson & Value Over Replacement Player (VORP).
  • Subsets:
    • player_match_vorp: 9,708 rows evaluating match contribution against replacement level (Surplus ratio: 0.997).
    • batting_predictions: 9,708 rows of forecasted runs with 50%, 80%, 90%, 95% intervals.
    • bowling_predictions: 9,708 rows of forecasted wickets and runs conceded.
    • fielding_predictions: 9,708 rows of expected fielding run values.
    • evaluation_metrics: Calibration diagnostics and benchmark comparisons.

3. Intended Uses

In-Scope Uses

  • Quantitative cricket analytics research and econometric modeling.
  • Point-in-time machine learning feature engineering benchmarks without lookahead bias.
  • Bayesian dynamic rating system validation and state-space estimation.
  • Empirical valuation of player marginal utility in limited-overs and multi-day cricket.
  • Reproducible academic demonstration and coursework in data engineering.

Out-of-Scope & Prohibited Uses

  • Live, in-play sports betting, gambling syndicates, or real-time odds calculation.
  • Commercial automated contract negotiations without contextual human audit.
  • Treating predictive outputs as guarantees of future sporting performance.

4. Data Collection, Ingestion & Quality

All underlying match data is sourced from Cricsheet (https://cricsheet.org). Cricsheet compiles ball-by-ball match scorecards published by cricket governing bodies and major sports broadcasters.

Ingestion Pipeline

  1. Raw Retrieval: ZIP archives (tests_male_json.zip, odis_male_json.zip, t20s_male_json.zip, ipl_male_json.zip, sma_male_json.zip) are downloaded and verified against cryptographic SHA-256 manifests.
  2. Quarantine Gate: JSON files are strictly schema-validated. Any malformed files missing core delivery structures are moved to data/quarantine/.
  3. Canonical Normalization: Deeply nested JSON arrays are parsed into normalized Third Normal Form (3NF) relational tables in DuckDB.
  4. Identity Disambiguation: 7,029 source player strings are linked to canonical player identifiers in dim_player_master using cross-referenced registry tokens with 99.98% verified confidence.

5. Known Biases and Limitations

  1. Domestic Tournament Coverage: Cricsheet provides complete coverage for the Indian Senior T20 domestic championship (Syed Mushtaq Ali Trophy), but does not provide ball-by-ball structured data for Indian First-Class (Ranji Trophy) or List-A (Vijay Hazare Trophy) competitions.
  2. Optical Tracking Data Absence: The dataset does not contain optical tracking coordinates (e.g., Hawkeye pitch trajectory, ball release velocity, swing deviation, or fielder GPS movement).
  3. Associate T20I Sample Sizes: Lower-tier international matches often feature small sample sizes and higher variance in opposing team quality.
  4. Historical Completeness: Comprehensive ball-by-ball data coverage begins around 2002 for international cricket and 2008 for franchise T20 leagues. Earlier historical eras are not represented.

6. Licensing and Attribution

This dataset is distributed under the terms of open data licenses:

Downloads last month
39