
Every dot is a brawler, every arrow a counter — it points from the winner to its victim. Colour marks the ones running in a circle: A beats B, B beats C, and C beats A again. No better brawler helps against a circle like that, because there is none. An arrow needs 150 duels of the same pair — the same bar as the counter list on his brawler page. So everything there is here.
These are always numbers from ranked. In trophy games it is not fixed who plays against whom, and without fixed line-ups “A beats B” cannot be counted. The other pages still show trophies.
101 brawlers, 676 counters in Hot Zone — holding 5 knots across 17 brawlers, in 33 arrows that run in a circle.
Charlie gets it worst: Kaze beats him far more clearly than the two brawlers' strength explains — across 670 duels, and the same in 4 modes.
The raw duel win rate is no good for this. It mostly measures who is generally strong — and a pure strength ranking cannot form a circle at all. Measured on 2026-08-20: among 32,447 pairings with duels at 60 % or above there are exactly zero circles.
So an arrow stands for the excess: the two brawlers' overall strengths give an expected rate, and whatever sits above it is the part strength does not explain. It is drawn from 5 points of excess and 300 duels — and only if the counter holds in at least two modes. Counter, hard counter and no chance split them in three: hard is the top quarter of all counters (today from 8.3 points), no chance the top five per cent (from 13.0). Both bars are shares, not fixed numbers — they are taken from all 2,015 arrows in the field, because fixed ones let more and more through as a season runs on. A circle must additionally close within a single mode; three arrows from three modes are not a circle anyone ever played.
The counter-check belongs here, because circles also arise on their own: a field of random duel values is full of rings. Across 30 runs with randomly shuffled excesses the average was 1.1 circles across all modes (0 to 4); measured are 68. So the existence of circles says nothing — that there are thirty times as many as by chance does. What makes the difference is the condition that every edge must hold in two modes: without it this would read 176 measured against 109 random, and that would be barely anything at all.
What is not here: whether a counter holds on every map. The duel numbers come per mode, not per map — map-dependent counters average out inside them.