# Prediction of the outcome of a Twenty-20 Cricket Match : A Machine Learning Approach

**Authors:** Ashish V Shenoy, Arjun Singhvi, Shruthi Racha, Srinivas Tunuguntla  
**arXiv:** [2209.06346](https://arxiv.org/abs/2209.06346)  
**Date:** 2022-09-13  
**Summary page:** https://www.summarizepaper.com/en/paper/2209.06346v2/

## Key points

- Twenty20 cricket, also known as T20, is a condensed version of the sport that has gained immense popularity in recent times
- Each team consists of 11 players who have a single innings each, limited to a maximum of 20 overs
- The shorter format was introduced to attract spectators and viewers by creating a more exciting and time-efficient game
- The research paper explores four different machine learning approaches for predicting T20 cricket match outcomes
- Factors considered include previous performance statistics of players, ratings from reputable cricket statistics websites, clustering players with similar performance statistics, and a novel ELO-based approach to rate players
- Different machine learning algorithms such as logistic regression, support vector machines, bayes network, decision tree, and random forest are used to evaluate the effectiveness of these feature engineering approaches
- A typical Twenty20 game lasts around three hours with each innings lasting approximately 75-90 minutes and includes a 10-20 minute interval
- Twenty20 cricket has been widely adopted around the world and has given rise to premier cricket league competitions like the Indian Premier League
- International tours usually include at least one Twenty20 match and domestic cup competitions are held by all Test-playing nations
- The International Cricket Council (ICC) aims to deliver real-time statistical information about players through platforms like the Cricket World Cup app or website
- The research paper aims to contribute to sports analytics by exploring different machine learning approaches for predicting T20 cricket match results.

## Summary

Twenty20 cricket, also known as T20, is a condensed version of the sport that has gained immense popularity in recent times due to its fast-paced and unpredictable nature. In a Twenty20 game, each team consists of 11 players who have a single innings each, limited to a maximum of 20 overs. This shorter format was introduced to attract spectators at the ground and viewers on television by creating a more exciting and time-efficient game. In this research paper, the authors explore four different machine learning approaches for predicting the outcomes of T20 cricket matches. They consider various factors such as previous performance statistics of the players involved in the competing teams, ratings obtained from reputable cricket statistics websites, clustering players with similar performance statistics, and propose a novel method using an ELO-based approach to rate players. To evaluate the effectiveness of these feature engineering approaches, the authors compare their performances using different machine learning algorithms including logistic regression, support vector machines, bayes network, decision tree and random forest. By analyzing these approaches and their respective algorithms they aim to provide insights into predicting T20 cricket match results accurately. The introduction section highlights that a typical Twenty20 game lasts around three hours with each innings lasting approximately 75-90 minutes and includes a 10-20 minute interval. This makes it much shorter than traditional forms of cricket and closer in duration to other popular team sports. The success of Twenty20 cricket has led to its widespread adoption around the world and has given rise to premier cricket league competitions like the Indian Premier League. Almost all international tours include at least one Twenty20 match and domestic cup competitions are held by all Test-playing nations. Furthermore, it is mentioned that one of the main objectives of the International Cricket Council (ICC) is to deliver real-time and interesting statistical information about players through platforms like the Cricket World Cup app or website. Fans are particularly interested in player performance statistics; therefore providing accurate information on each player's performance is a priority for both ICCs broadcasting channels alike. Overall this research paper aims to contribute to the field of sports analytics by exploring different machine learning approaches for predicting T20 cricket match results.