# An Application of Logistic Regression

Here are some reactions to our latest analysis. Can you guess what the topic is?

“This is awesome. Fascinating work! I always knew this intuitively but you gave it the statistical backbone. Superb work!”
“OH NO. STOP THIS MEANING LESS STATISTICS”
“perfect analysis bro…keep it up.”
“Absolutely rubbish analysis… “

Full marks to you if you guessed it. Yes its about Cricket.

Nothing gets Indians excited like Cricket. I recently wrote an article about who is the best ODI batsman for India. The analysis is fairly simple. I compared India’s top 3 all-time scorers i.e. Sachin Tendulkar, Rahul Dravid and Sourav Ganguly.

I built a model to calculate the impact of each run scored by these 3 batsmen on the team’s chances of winning. I then used it to calculate the average impact per inning. It turns out that in every inning that Dada has batted for India, he has improved the team’s chances of winning by 13% on average. This can be called Dada’s contribution from his batting in every inning.

Similarly, Dravid’s contribution is 11% and Sachin’s contribution is 10%.

Thus it turns out that Dada has had a greater impact on the team’s win rate than both Sachin and Dravid.

As an analyst, I know that this is just one way of looking at things. For example, if instead of using averages, we use the totals, then we can calculate the life time contribution of each batsman. (We have calculated average contribution per inning. we can also calculate the total lifetime contribution.)

There are dozens of other variables that can be added to make this analysis more robust. Is a victory against Australia or Pakistan more valuable than a victory against Bangladesh. If it is, then we need to add this information to the model. As I said, this is a very simple 1-variable regression analysis.