GROUP 2 - DSBA

Batch:2023 Bengaluru

Abdul Haseeb Pasha –

Arun Balaji - Basavaraj
KY – Dylan Furtado –
Harsha MS – Mittal
Shah – Nikhil
Upadyaya
Research Supervisor – Nimes
Synopsis of Player Valuation in Indian Premier League Auctions
Contents
1. Introduction..................................................................................................................................2
2. Scope & Objectives.......................................................................................................................2
3. Data Sources and Description.......................................................................................................2
4. Analytical Approach......................................................................................................................3
5. Recommendations & Applications................................................................................................3
6. References and Bibliography........................................................................................................3

1|Page
1. Introduction
The Indian Premier League (IPL) stands as a groundbreaking cricket league that has captured
global attention since its inception in 2008. Founded by the Board of Control for Cricket in
India (BCCI), the IPL has redefined the sport with its fast-paced Twenty20 (T20) format, star-
studded lineups, and innovative approach to entertainment and commercialization. Over the
years, it has become a symbol of sporting excellence, attracting millions of fans worldwide
and revolutionizing the cricketing landscape. Moreover, the role of data science has emerged
as a crucial factor in the league's success, transforming player recruitment, strategy
formulation, and fan engagement. Data science drives decisions in player auctions,
performance analysis, opponent scouting, and even enhances fan experiences through
personalized engagements. With millions of data points analyzed each season, data science
has become the MVP (Most Valuable Player) of IPL's playbook, shaping the game both on
and off the field. Let's explore into how data science, with its numbers, has helped the IPL
franchises and its impact on the game.

2. Scope & Objectives

The scope of this capstone project encompasses the development of a Machine Learning
model tailored for IPL franchise owners to optimize their player bidding strategies during
auctions. The model will utilize diverse data sources, including historical player performance,
team strategies, and probability of winning, to inform decision-making processes.
Additionally, social media analytics will be integrated into the model to explore alternative
revenue streams associated with player acquisitions.

Objectives:
 Develop a robust Machine Learning model that leverages past player performance,
team strategies, and winning probabilities to optimize player bidding strategies
during IPL auctions.
 Incorporate social media analytics into the model to analyze factors such as social
media presence, engagement metrics, follower demographics, and sponsorship
potential of players.
 Provide actionable insights to IPL franchise owners, enabling them to maximize player
value and capitalize on potential revenue streams beyond on-field performance.
 Empower franchise owners with the tools necessary to make informed decisions
during player auctions, ultimately enhancing their competitiveness in the IPL
ecosystem.

2|Page
  Enhance the understanding of the value proposition associated with player
acquisitions by considering both on-field performance and off-field revenue
potential.

3. Data Sources and Description

Data Sources

IPL Data Science Article: This article from Analytics Vidhya provides insights into the data
science techniques used in analyzing IPL data. While it doesn't directly provide datasets, it
offers valuable information on the types of analysis and methodologies relevant to the IPL
problem statement.

IPL Matches Dataset (2008-2020): This dataset is hosted on Data.World and contains detailed
statistics of IPL matches from 2008 to 2020. It includes information such as match outcomes,
player performances, team strategies, and more, which can be valuable for building
predictive models and analyzing player auctions.

Sports Datasets on Data.gov: Data.gov offers a collection of sports-related datasets that may
include relevant information for the IPL problem statement. Exploring this repository can
potentially uncover additional datasets related to player statistics, team performance, and
other relevant factors impacting IPL auctions.

Google Dataset Search: Google Dataset Search provides a platform to discover a wide range
of datasets from various sources across the web. Searching for sports-related datasets using
keywords such as "cricket," "IPL," or "sports analytics" may yield additional datasets suitable
for analysis and model development.

4. Analytical Approach

Data Collection

Historical Player Performance:

Runs scored, wickets taken, batting average, bowling average, strike rate, economy rate,
number of catches taken, etc. This data can be obtained from various cricket websites or
APIs.

Team Strategies:
Winning records, playing history (home/away), team composition (balance between
batsmen, bowlers, all-rounders), etc. This data can be obtained from the IPL website or
official scorecards.

Social Media Analytics:

3|Page
Social media following (number of followers on Twitter, Instagram, etc.), engagement metrics
(likes, shares, comments), follower demographics (location, age, gender), brand
endorsements. Social media APIs can be used to collect this data.
Data Preprocessing

Handle missing values:

Some data points might be missing. You can impute them using mean/median values or
remove rows with missing data.
Feature scaling: Since different features will have different units (runs, wickets, followers etc.),
scaling them to a common range will improve the model’s performance.
Exploratory Data Analysis (EDA)

Analyze the distribution of player prices to identify outliers and potential trends.
Visualize the correlation between player performance metrics (runs, wickets) and price to
understand which factors have a stronger influence.
Use techniques like Principal Component Analysis (PCA) to identify underlying patterns in the
data and reduce dimensionality.

Model Building

Machine Learning Algorithms: The image you sent shows various Python libraries and
techniques commonly used in machine learning, including:
Linear Regression: This is a good starting point for modelling continuous variables. It can
help you identify the linear relationship between features (e.g., runs scored) and player price.
Random Forest Regression: This is a robust ensemble technique that can handle complex
non-linear relationships and works well with high dimensional data.
Other algorithms you can consider include K-Nearest Neighbors (KNN), Support Vector
Regression (SVR), and Gradient Boosting Regression.

Model Evaluation

Split your data into training and testing sets. Train the model on the training set and evaluate
its performance on the testing set using metrics like Root Mean Squared Error (RMSE) or
Mean Absolute Error (MAE).
Fine-tune the model hyperparameters to improve its accuracy.

Social Media Analytics Integration

Develop a separate model to predict a player’s brand value/sponsorship potential based on

social media engagement metrics and follower demographics. This can be integrated with
the main player price prediction model to provide a holistic view of a player’s value.
Additional Considerations

4|Page
Team Budget Constraints: The model should incorporate a team’s budget constraint to
ensure the predicted price stays within realistic limits.
IPL Auction Dynamics: The IPL auction is a dynamic event where bidding wars can inflate
prices. The model should account for this by incorporating an element of uncertainty or
randomness.

By following this approach, you can develop a machine learning model that provides
valuable insights to IPL franchise owners to optimize their bidding strategies during auctions.

5. Recommendations & Applications

Optimizing Player Bidding Strategies during IPL Auctions:
Utilize the developed Machine Learning model to analyse historical player performances,
team strategies, and winning probabilities to inform IPL franchise owners' decision-making
processes during player auctions.
Incorporate the model's insights to strategically bid on players who align with team
requirements, budget constraints, and have high potential for on-field performance and off-
field revenue generation.

Maximizing Revenue Streams Beyond On-field Performance:

Identify players with high brand value and sponsorship potential through social media
analytics.
Empower IPL franchise owners to capitalize on off-field revenue streams by strategically
acquiring players with significant social media presence and engagement, leading to
increased sponsorship opportunities and brand endorsements.

Improving Team Competitiveness and Performance:

Provide actionable insights derived from the Machine Learning model to IPL franchise
owners to optimize player selection, team composition, and strategy formulation.
Enable franchise owners to make informed decisions that enhance their team's
competitiveness in the IPL ecosystem, leading to improve on-field performance and
increased chances of success in the league.

By implementing these recommendations and applications, IPL franchise owners can

leverage data science to optimize their player bidding strategies, enhance fan engagement,
maximize revenue streams, improve team competitiveness, and contribute to the ongoing
evolution and success of the Indian Premier League.

6. References and Bibliography

<Details about any reference books, articles, web resources etc. that are to be used for
the project. Since students are expected to do their Capstone Project as a research thesis
and not like an assignment, write at least ten references or links to the references.>

5|Page