Families and their incomes. Students and their test scores. Children and their heights.
Chances are that you have begun your journey into the world of econometrics by working on case studies of this sort. Which is all well and good, but what if you could learn econometrics by analyzing cricket matches?
If you’re someone like me, you’re likely salivating at the very thought.
Misra’s cricket connections provided the key to the “learning” that would fuel their new model: data from 4700 ODI matches. Runs and wickets counted as sequential data, balls bowled would represent “time” through which the score progressed, the “intervention” would be something like inclement weather. The counterfactual question: if you stopped an innings right now, how would the trajectory of run-scoring proceed?https://fiftytwo.in/story/numbers-game/
I spend most of time thinking about motivating questions. I don’t know if the term is academically accurate, but to me, it simply means this: what question is likely to motivate a student to want to know more?
To me, these are good questions to begin classes on macroeconomics, hypothesis testing and development economics, because they make the student curious to know more. “I need to know this in order to score marks in an exam I have to give” is bad motivation. Because all that the student is going to do is minimize their effort in order to maximize their marks.
But “this is fascinating! Tell me more!” is such great motivation! Asking a motivating question is half the battle done, because you’re likely to push the student off on their own trip. And that is A Very Very Good Thing! The technical term for this, by the way, is heutagogy.
And “Can you build a model that will beat other student’s models in terms of accurately forecasting tonight’s IPL matches scores” is an excellent motivating question.
Imagine a summer school that started on the first day of the IPL this year, with a very simple objective. The student whose prediction for the IPL final’s scores turned out to be the most accurate would win a jackpot prize, and you have the duration of the tournament to figure out how to build such a model.
That’s the “syllabus” for this course, and also the learning objective. There’s no textbook, no fixed course material, no “lectures”. Spend six hours everyday figuring this out, and when you are stuck, you can speak to the faculty, who might recommend a particular topic to look up online. You learn as you go along, and you have help you can count on.
You would end up learning about least squares, gradient boosting, decision trees, nearest neighbours algorithms, and so much more. You would also need to learn about webscraping, coding, iterating and refining your model.
Statistics, and boring? You must be out of your mind!
If you are a student reading this, please read the rest of the article, and try reaching out to some of the folks mentioned in it. Ask if they would be willing to give a lecture (online/offline) in your college. Make a note of the firms mentioned, reach out to them and ask the same thing. Best of all, speak to your placement cell and ask if it might be possible to get these firms on campus for recruitment.
And above all, do let me know if your model is getting good at predicting scores in matches. It’ll help me plan my time better during the upcoming T20 World Cup! 😉