What is Criclytics?
Published · July 1, 2022
Criclytics is a predictive analytics system built for cricket. It is the engine powering the predictive layer of cricket.com, a flagship product of Head Digital Works Pvt. Ltd.
It uses historical match data, statistical models, and machine learning to answer one core question at different stages of a match:
What is likely to happen next?
Criclytics works across T20, ODI, Test, and T10 formats, for both international and domestic cricket. It covers the full journey of a match:
- Before the match starts
- While the match is live
- After the match ends
Over time, Criclytics expanded into multiple supporting engines that improved fan engagement, content discovery, and social reach.
Criclytics (Pre, Live and Post Match)
Criclytics is built around three stages of a cricket match.
Pre-match Predictions
Before a match begins, Criclytics estimates:
- Team win probability
- Projected team score
- Player runs and wickets
- Key batter vs bowler matchups
- Team strength using a player and team rating system
These predictions are based on:
- Historical performances
- Player form
- Head-to-head history
- Rating parameters for teams and players
Live-match Predictions
During the match, Criclytics updates predictions ball by ball.
It uses:
- Current score
- Strike rate
- Bowling average
- Balls remaining
- Target score
- Pre-computed team ratings
Machine learning classification and regression models keep updating:
- Win probability
- Projected runs
- Projected wickets
- Projected overs
A Monte Carlo simulation layer runs in the background to simulate:
- Player runs
- Wickets
- Partnerships
- Tie probability
This allows Criclytics to capture changing match situations with high sensitivity.
Post-match Summary
After the match ends, Criclytics generates:
- Match stats
- Scoring charts
- Phases of play
- Match reel summary
This helps users quickly understand how the match evolved over time.
FRC (Fantasy Research Center)
FRC was built as an extension of Criclytics for fantasy cricket users.
Using the same prediction and rating backbone, FRC helps users:
- Understand key player matchups
- Evaluate player form
- Build better fantasy teams using data
- Identify high impact players for a match
Article Reco Engine
To improve content discovery, a simple but effective article recommendation engine was built.
When a user finishes reading an article, the system suggests similar articles.
This engine uses:
- TF-IDF vectors to represent article content
- Cosine similarity to find related articles
- A content-based recommendation approach
This significantly increased the average time users spent on the app.
Twitter Auto Posts
Criclytics insights were also used to automatically generate and post images on Twitter at the right moments during a match.
These posts included:
- Key batter vs bowler matchups
- Toss update and win prediction
- Playing 11 announcement
- Fantasy best team
- Best player of the match
The images and tweets were generated programmatically and posted faster than manual updates, increasing engagement and traffic back to the platform.
Challenges
Building Criclytics was not only about models. Some key challenges were:
- Handling live match data updates reliably
- Keeping predictions fast enough for real-time use
- Making complex predictions understandable to users
- Scaling the system across formats and matches
- Ensuring consistency between pre, live, and post match engines
Criclytics showed how data science can be applied to a fast moving sport like cricket in a practical and engaging way. It combined predictive models, simulations, content systems, and automation into one connected product experience.