I Trained for a Half Marathon While Doing My B.Tech Thesis. Here Is the Data.

The rational case for training for a half marathon during your B.Tech thesis semester is weak. The sleep debt alone makes the decision questionable. I have the data, because I tracked everything in a spreadsheet because I am that kind of person, and the data is humbling. Average sleep: 5.2 hours per night over 14 weeks. Average weekly mileage: 35 kilometres, which is a moderate plan for a beginner and a heroic plan for someone also writing a thesis.

I did it anyway and I finished in 2 hours and 4 minutes at the IIT Madras 5K/10K/HM event in February, and it is one of the things I am most proud of from my undergraduate, which tells you something about either the experience or my other accomplishments, possibly both.

Why I Did It

My thesis was about adversarial robustness in neural networks. It was going adequately but not brilliantly. I was spending twelve hours a day staring at training curves that did not improve as fast as I wanted them to. I had been running casually for a year. In October, someone on my hostel floor mentioned they had signed up for the HM and I signed up on an impulse that I did not fully interrogate until I was five weeks into a training plan and already sleep-deprived.

The reason the impulse made sense, when I examined it: the thesis was a project where progress was invisible and slow. A training plan has workouts. A workout either happens or it does not. The half marathon either happens or it does not. The clarity of having a thing that was either done or undone was, during thesis season, enormously psychologically useful.

The Spreadsheet

I tracked: sleep (hours), morning heart rate (BPM), thesis words written, running kilometres, and a subjective 1-5 score for how the day went. Patterns I observed: days after long runs were consistently higher productivity thesis days, despite lower sleep. Days I skipped running were inconsistently productive. The best thesis weeks coincided with mid-volume running weeks, not rest weeks. I am aware this is n=1 data from a biased experimenter. It is still data.

The Race

Race morning: 5:30 AM start, 17 degrees, the IIT Madras research park loop, which is flat enough to be kind to someone who trained mostly on flat roads. I ran with a playlist and without a watch, which was a decision I made at 11 PM the night before and which I recommend. Crossing the finish line at 2:04 was the first time in months that something I had been working on had a clear endpoint that I reached. My thesis needed more work. My legs had done the thing they had been asked to do. I sat on the grass for twenty minutes and felt completely fine.