I had the opportunity to participate in a Datathon this Friday, organized by Correlation One for Citadel Asset Management (and Citadel Securities). It was an incredible 8-hour sprint of data science and statistics, and -though I can still hardly believe it- our team walked away with first place, and the $20,000 prize.
I’ve organized some Hackathons in the past, but this was different: rather than pitching interactive software to investors, we were tasked with writing a report showcasing interesting insights mined from the datasets we were given. While the hackathons I’ve been in have been extremely stressful as I try to build code that plays well with others, the Datathon gave me the same level of focus but without the stress. For me, this felt like another day in the lab- working with iPython to structure a dataset, identify correlations, test hypotheses, and write up the results.
While I can’t share any details about the datathon topic and prompt, I have some general thoughts on how to succeed. Note that the format may have changed since I participated!
Before the Datathon:
- Decide if you’re a programmer, or a data scientist. The screening questions will be very different if you’re a ‘researcher’ (i.e. statistician, data scientist, mathematician) or an ‘engineer’ (i.e. good at database optimization and production C++ code).
- Take the screening assessment seriously: I was expecting a simple Eventbrite signup; instead I spent 90 minutes doing math and programming out technical questions. Take it seriously, but have fun.
- Find a team during the networking session: I hand’t known any of my teammates before the event, but we had a great time and look forwards to working together in the future. Be open about your skills, abilities, and whether you’re still looking for teammates.
During the day of the Datathon:
- Don’t just plot the data: We saw many teams who just did exploratory data analysis, plotting trends and covariance matrices. This doesn’t add any value to the dataset- what can you predict from the data; what new hypotheses can you explore?
- Tell a story: The winning teams pitched topics of interest to policymakers, businesses, or consumers. How can your insights be turned into real actions?
- Test many hypotheses: Our team tested and threw out ~4 hypotheses before choosing one to focus on.
- Begin With the End in Sight: Because we didn’t have the chance to pitch our idea in-person, our final report had to speak for itself. Make it professional, clear, and concise.
It’s been a great boost to my resume to add the Datathon win… and a nice bump to my bank account!