Data-Driven Cricket Player Selection Using Logistic Regression: An Explainable Framework for Playing XI Prediction and Player Ranking

Selecting a cricket team involves more than choosing the players with the most runs or wickets: role, recent participation and team balance all matter. This paper proposes an explainable, leakage-free framework that predicts the Playing XI of an IPL team before a match using Logistic Regression and role-constrained integer optimisation, and measures the prediction against the XI that actually played. The study uses the IPL Complete Dataset (2008–2024), with actual XIs taken from Cricsheet. For each upcoming match, features such as batting and bowling statistics, decay-weighted recent form, consistency and recent participation for the team are computed only from earlier matches. Logistic Regression estimates, for every player in the team's candidate pool, the probability of being in the XI; an integer programme then selects eleven players under wicketkeeper, batter, bowler and all-rounder constraints. Chronological splits are used: training up to 2021, validation in 2022–2023 and testing in 2024. On the 2024 test season the model reached a ROC-AUC of 0.940, accuracy of 87.9% (majority-class reference 63.3%) and a Brier score of 0.091. In a walk-forward backtest over 71 matches (142 team–matches), the constrained XI contained on average 8.57 of the 11 actual players (77.9%; median 9; range 4–11), against a ceiling of 10.47 set by the candidate pools and 6.44 for a career-totals baseline. Two simple baselines, the unconstrained top 11 (8.95) and repeating the previous XI (8.96), were slightly better than the proposed method, so it did not exceed a simple continuity rule; enforcing role constraints cost about 0.4 hits. The strongest predictors were recent participation for the team and recent form. Limitations include a single test season, derived candidate pools and a manually built role map.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24322163
Primary Topic
Sports Analytics and Performance
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Data-Driven Cricket Player Selection Using Logistic Regression: An Explainable Framework for Playing XI Prediction and Player Ranking

Aryan Kaushik, Aryaveer Sharma
2 links
Zenodo (CERN European Organization for Nuclear Research)
Sports Analytics and Performance
preprint

Data-Driven Cricket Player Selection Using Logistic Regression: An Explainable Framework for Playing XI Prediction and Player Ranking

Aryan Kaushik, Aryaveer Sharma
preprint en

Abstract

Selecting a cricket team involves more than choosing the players with the most runs or wickets: role, recent participation and team balance all matter. This paper proposes an explainable, leakage-free framework that predicts the Playing XI of an IPL team before a match using Logistic Regression and role-constrained integer optimisation, and measures the prediction against the XI that actually played. The study uses the IPL Complete Dataset (2008–2024), with actual XIs taken from Cricsheet. For each upcoming match, features such as batting and bowling statistics, decay-weighted recent form, consistency and recent participation for the team are computed only from earlier matches. Logistic Regression estimates, for every player in the team's candidate pool, the probability of being in the XI; an integer programme then selects eleven players under wicketkeeper, batter, bowler and all-rounder constraints. Chronological splits are used: training up to 2021, validation in 2022–2023 and testing in 2024. On the 2024 test season the model reached a ROC-AUC of 0.940, accuracy of 87.9% (majority-class reference 63.3%) and a Brier score of 0.091. In a walk-forward backtest over 71 matches (142 team–matches), the constrained XI contained on average 8.57 of the 11 actual players (77.9%; median 9; range 4–11), against a ceiling of 10.47 set by the candidate pools and 6.44 for a career-totals baseline. Two simple baselines, the unconstrained top 11 (8.95) and repeating the previous XI (8.96), were slightly better than the proposed method, so it did not exceed a simple continuity rule; enforcing role constraints cost about 0.4 hits. The strongest predictors were recent participation for the team and recent form. Limitations include a single test season, derived candidate pools and a manually built role map.

Zenodo (CERN European Organization for Nuclear Research)
The NorthCap University (IN)
Sports Analytics and Performance
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon