JardineAI
It is a cricket related project which aims at cloning the behavior of a batsman, letting the machine learn to bat like any batsman through his previous batting career and predict the batsman?s complete response against any delivery under any playing conditions including the shot selection, footwork
2025-06-28 16:33:55 - Adil Khan
JardineAI
Project Area of Specialization Artificial IntelligenceProject SummaryIt is a cricket related project which aims at cloning the behavior of a batsman, letting the machine learn to bat like any batsman through his previous batting career and predict the batsman’s complete response against any delivery under any playing conditions including the shot selection, footwork and the most probable outcome against the ball etc.
The system would identify and extract any batsman’s weaknesses and strengths and propose areas and deliveries that can be most effective against that batsman. Letting a team pick best team combinations and develop strategies and tactics and revolutionizing cricket training, statistics and analysis.
- Collecting and compiling data regarding all the balls faced by a batsman, i.e thier line, length, pace etc and batsman's response against it.
- Formatting the data for Supervised Machine Learning.
- Training formatted data of the batsman with Neural Networks to prepare any batsman's clone.
- Cloning complete behavior of the batsman.
- Taking predictions from the clone and analysing them.
- Calculating the strength and weaknesses of the batsman.
- Data Visualization to let experts, trainers, managers and players observe strength and weaknesses of opponents or own players.
First we'll collect the data of all the balls that batsman has played in his career, the data will be formatted in a specific order so that it can be used for supervised machine learning. The formatted data will be modeled and trained with some neural networks and the trained datasets will be stored in our database.
Now clone takes a delivery and using trained datasets, gives a response which that batsman will most likely give.
For that we have multiple classifiers, one predicts the outcome (like a single or a boundary or wicket etc), the second classifier predicts the batsman will respond on which foot (like on front-foot or back-foot etc). Third classifier predicts the direction of the face of the bat, which gives us the stroke that batsman will most probably play (like if the direction is 45 degrees, it's a cover drive) and so on.
This way, we'll have a complete clone of that batsman, responding almost like that batsman might.
Lastly, an Analyzer takes predictions against all the deliveries and sorts and filters it out to prepare this final output which will tell us what sort of balls are predicted to trouble the batsman most or what balls are supposed to get hit away. So we'll know what to bowl to that batsman and what not to. Thier will be detailed behvior logs, behavior graphs and vulnerability visualizations etc
- Know strong zones and weak zones of any opponent batsman,
- Exploit falws in opponents batting techniques,
- Explore personal strength and weaknesses,
- Train and work to overcome detected weaknesses,
- Prepare most effective strategies and tactics,
- Develop better gameplans,
- Get detailed list of balls with highest probability of getting an opponent out,
- Discover new traps and strategies,
- Virtually train against any batsman's clone,
- Pick best possible line ups and bowling combinations considering opponents Jardined results
...etc
Final deliverable will be a software system, it will be a web based system and will be built primarily in Nodejs enviroment, Javascript being core language.
Tensorflow.js will be used for most of the machine learning.
Modelling, training and all other preparation of data sets will use Neural Networks.
Data visualization will be done mainly with Plot.ly and chart.js.
Ball by ball data collection will be implemented through web scrapping using Puppeteer.
If a real-time ball tracking system is build and embedded for live dataset generation, it will be developed with Tensorflow.js, opencv.js and tracking.js.
Interactor will be able to select any out of the cloned batsmen and bowl at him, in response he/she'll get the predicted response of that batsman and most probable outcome against the bowled ball.
The system will be able to play against the clone itself and detect the clone's strong and weak areas and output the most effective and least effective balls against the batsman and plot pitch maps and other behavior graphs to let the interactor analyse the behavior of the batsman.