Who to pick, who to leave and who are the Best playing 11 for a T20 match? This blog gives ideas around who to pick from the cricket teams for your fantasy league team.
With the growing craze for T20 cricket, the smallest format of the game, and the introduction of Indian fantasy sports platforms such as Dream 11, we thought what if we could use data to analyze the sport and get an upper hand while playing on these fantasy sports platforms. So, we decided to select the best 5 batsmen, 5 bowlers, and 1 wicketkeeper from the playing 11 of both the teams combined based on their past records in the IPL.
Data Collection
We begun the process by first collecting the data that we scraped from the website: https://stats.espncricinfo.com/ci/engine/stats/index.html?sdb=find and also used the data from https://cricsheet.org/. We collected the data of all the matches of IPL from 2008 to 2021.
Feature Engineering
By using the ball by ball data obtained above we created a database for the stats of every player which contained features such as Runs Scored, Strike rate, Boundaries Scored, performance in the powerplays and the death overs, and their performance against a particular team. These are few of the different parameters which can be used for the prediction of best 11 players in a particular match.
Modeling and Weight Distribution
We have now created a model to get a performance score based on the above criteria and selected the players from this with the highest performance score.
As we know that the relation between the performance of a player and the different features is not the same so we give them different weights (percent contribution made by that feature in the calculation of performance score) like
For batting -> runs a weight of 8.63, strike rate 7.62, dot ball percentage 1.65, fours 5.24, sixes 5.70, etc.
For bowling -> wickets a weight of 12.91, dot balls bowled 8.09, runs conceded 2.81, extras 2.78, etc.
For wicketkeeper -> we decide based on the no of stumpings and catches.
We know to see the relation of the performance score calculated above with various features, plotted the below few graphs:
Plot for Performance score vs feature - runs 2008 to 2018
Plot for Performance score vs feature - fours 2008 to 2018
Plot for Performance score vs feature - sixes 2008 to 2018
Validation of model on new match
We analyzed our results using a match between Mumbai Indians and Chennai Super Kings which took place in Pune on 28/04/2018 so we selected the players using our model and data from 2008 to 27 Apr 2018.
Here we predicted the top 5 batsmen for the match and checked it with their performance on this match:
Here we can see that we correctly predicted 3 out of the top 5 batsmen and 1 wicketkeeper.
We will now predict the top 5 bowlers for the match and checked their performance on this match:
Here we can see that we correctly predicted 4 out of the top 5 bowlers.
In the overall match, we were able to predict 8 out of 11 players accurately. This shows that there is a relation between a player’s performance and his records.
The model was able to detect it up to a good extent with more than 70% of accuracy. More accuracy can be achieved by taking into the factor of new players who have not played IPL before but have data from another cricket tournament. We will look into it in the next article.
Thank you for reading, we hope you find our article interesting!
If you have any questions about how to use these data insights in your business and engage your customer segment then contact us here - contact@godatainsights.com.
Author-