Column
Articles
Data-driven ARAM Mayhem analysis written from real matches: champion identity, augment trade-offs and the statistics behind the tier list.
Published: 2026-07-07
Every article here starts from the ARAM Mayhem matches arammeta collects. Each one defines its metric first, then shows the ranking and the counter-examples, and closes with the data range and its limits. Numbers carry their sample and patch; they describe associations in past games, not a guarantee about any single match.
- Power curve
Kill sprees do not equal snowballing
Correlation between the kill-spree axis and actually closing games early: rho = −0.03, essentially zero. The fastest closers are Yorick and a row of zero-spree supports, while spree kings like Twitch and Zed win slow.
- Skill-scaling
Skill-scaling: who rewards skill, who farms low elo
The same champion, split by lobby skill: Vayne climbs +5.0pp in strong lobbies, Sion drops −4.3pp. Who rewards good play, and who just farms weak games.
- Champion map
Late game × snowball: a champion map
Every champion plotted on two axes — snowball potential and late-game strength — so you can see who rushes, who scales, and who does both.
- Augment deep-dive
Draw Your Sword: who abuses it, and how to build
Ranked by lift (champion strength removed): Samira tops at +20.7pp, plus the data-backed core build.
- Basics
How to read this tier list
How win rate, Bayesian shrinkage, sample size and tier thresholds are computed and read — and what the numbers cannot tell you.