SportsKPI has been one of the pioneers when it comes to Kabaddi analytics in India. A deep dive into the sport helped us formulate the True Raiding Impact (TRI), Net Points per Raid (NPpR), and Net Points per Tackle (NPpT) models.

These helped to better evaluate players by analyzing their actual impact on the mat, as opposed to just assessing them based on raw numbers. As time has gone by, though, there has been a need for a newer and better model in Kabaddi. To fill that need, we teamed up with the Decision Sciences Department at IIM Bangalore and the results were terrific.

Key characteristics of the new model:

1. Improved Raid outcome prediction
2. Improved Win % prediction
3. A unique Raider Score

The model analyses the player’s past performance and comes up with the course of action for them on the mat that would lead to the best possible result. It also predicts how important a raid is toward winning the game. This, in turn, helps assess players in different in-game situations, i.e., how they fare when the pressure is high and when it is low. To create these improvements, this model was formed using logistic regression and included interaction variables to cover all cases.

The model also provides us with a Raider Score, which is derived after a comprehensive evaluation of the player in question. The players are then divided into three separate clusters, with the third one representing the best of the best.

The new model helps tremendously in developing an auction strategy by identifying who are truly the best players available. Then in the games, it assists in making better decisions in order to improve performance on the mat and the results.

Stay tuned for a more in-depth look in the future!