LCK — Spring 2021, Week 4

Introduction

I am a machine-learning engineer by trade, and a bad league of legends player in my free time. I have a passion for using statistics to solve problems, so last year I developed an algorithm for quantitatively scoring League of Legends pro players. I use this algorithm to rank players and combine the player scores to rank teams. I used this algorithm to correctly predict every series from Worlds 2020 (excepting SN over TES).

This year, I’m planning to iterate and improve my algorithm to see how accurate I can become in predicting game / match winners. Every week, I’ll post my rankings and predictions so that I can be on the record. I’ll also keep track of my performance as the season progresses. I’ll write up a brief summary of the top tier of teams for each region. I’ll also highlight any games that will be great to watch due to the power of the teams playing.

As always, you can check out the comprehensive rankings for all 4 major regions on my website.

Ranking Algorithm

The algorithm I use to calculate power rankings utilizes statistics from Oracle’s Elixir. To eliminate the variability caused by different game lengths, I use almost exclusively time-independent or time-standardized statistics. For example, I use statistics that are calculated “per minute” rather than “per game”, and statistics that are taken at standard times, like CS difference at 15 minutes.