Auquan · Spring Challenge

Auquan Spring Challenge

Bored one week, I entered a competition. A plain tuned model beat everything else I tried.

  • Python
  • Machine learning
  • Logistic regression
  • Hyperparameter tuning
  • Winner

One week in 2019 I was getting bored. I saw a competition (Auquan's Spring Challenge) and decided to spend some time on it. There wasn't much more to the decision than that. I entered, submitted something, and then went back to whatever else I was doing.

What I actually did

I was working through ML courses at the time, and I wanted to see whether any of what was being taught had a practical application. So I treated the challenge as a place to try things, without worrying much about where I'd place.

I tried a fair amount. In the end, what performed best was logistic regression with hyperparameter tuning, not the most exotic thing I reached for, just the thing that worked.

I more or less forgot about it after submitting. The judging took a while.

The email

Then, months later, an email arrived. I was one of the winning competitors. My submission had scored above 0.03, which the judges had decided put me in the top prize bracket, and there was a cash prize attached to it.

Email from Auquan confirming Shubhransh Jagota as one of the winning competitors of the Spring Challenge, scoring above 0.03 and placing in the top prize bracket
The email from Auquan, July 2019. Headers cropped and the prize amount blurred

The point

I got lucky in the end. But I only got lucky because I entered, and I very nearly didn't. It was a bored week and it would have been easier to do nothing.