We had recently visited Purdue for a hackathon. It was one of the most quietest towns we’ve been to and for the first time vivmbbs and naonao experienced snow! The temperature was in 20’s, freezing the hell out of us. But, nothing deterred us from doing the hackathon with some of the best CS brains in the world.
The hackathon started off with a talk about the company followed by a super interesting Game Prediction problem designed by naonao for the popular game League of legends. Given a training data set (a real one), you were supposed to predict which team wins based on the champions they’ve picked.
It was a 4-hour contest and a number of interesting algorithms came out of it. The most popular one was:
A basic approach to this challenge was to find the win percentage of each of the different heroes in the training set. Then you would predict which team would win based on the team with heroes with higher win percentages. This is the approach that most of the contestants took and yielded an accuracy of 60.8%. The top contestants did more and manage to squeeze out a little more accuracy. Since the data is real, that might be as accurate as it can get based just on the hero selection because there are many other factors that go into deciding which team will win.
We had a small glitch in the leaderboard which summed up the scores if you submitted twice exactly at the same time. How do you do it? by clicking on the button twice. Suddenly, jin75 was at the top with 112.12 points when the maximum score is only hundred. The leaderboard auto-refreshes itself every minute and it got automatically fixed.
It was a pretty intense competition at the top finally won by n000b. He’s a 16-year old prodigy, not finished high school yet but is already taking classes at Purdue, grabbed an intern at Facebook this summer and even a T.A as well!
Most of them coded in Python, Java & C++ with the exception of ertemplin who chose Go! There was a $512 prize pool with the winner getting $256 followed by $128, $64, …. $1
Special thanks to formatjam for helping us organize this hackathon. We really loved Purdue and would love to come again. Thanks again for participating in the contest! The challenge is still up for practice and you can go to hackerrank.com/purduehack
Interested in hosting a hackathon at your school? Please reach out to Andrew (andrew [at] hackerrank)
PS: The profile link hasn’t incorporated the hackathon scores yet. It’ll be done soon.