Predicting League of Legends Match Outcomes from Early Game Statistics

The Story Behind This Project

It all started in October 2017. I was a sixth grader back then, and I stumbled upon League of Legends esports for the first time. I became a die-hard fan of RNG, a legendary team from China. I still remember the thrill of watching them set up an early-game ambush, catching the enemy off guard and securing first blood within minutes. But even as a kid, I couldn't help but wonder: how much does that one kill actually matter? If my team is 2,000 gold ahead, what does that really mean for our chances of winning?

Eight years have passed since then. I went from being a spectator to a player myself, and those questions never left me. Every game I played, I found myself thinking: what really matters in the early game? What leads to victory?

Fast forward to college, where I finally got the chance to study machine learning systematically, especially feature engineering and predictive modeling. That's when it hit me: could these tools finally answer the questions I've been carrying since childhood? I found this Kaggle dataset, and here's the crazy part: it was created around the exact same time I first time playing League of Legends. It felt like the universe was telling me something. So here we are: a machine learning project eight years in the making, built to answer the questions of a curious sixth grader who just wanted to understand the game he loved.