Sniper

Nothing in this dataset measurably beats Sniper. Seven matchups clear the significance test in its favour — Lich at 64.0%, Zeus at 61.5%, Ursa at 61.1% — and not one clears against it. There is no hero you can point to and say: pick that, and Sniper has a problem.

And yet the win rate falls from 51.3% in Herald to 47.5% in Divine, dropping below fifty somewhere around Archon and continuing down. Something is beating Sniper. It just is not a hero.

This combination is close to unique here. Seven heroes have no reportable unfavourable matchup at all, and the other six all climb as brackets rise — Treant Protector, Batrider, Invoker, Keeper of the Light and Shadow Demon among them. Sniper is the only one that declines, and it declines by 3.8 points.

The natural reading is that Sniper’s weakness is not a counter but a competence. A matchup table can only measure things that come in the shape of an opposing hero. It cannot measure whether the enemy team walked down a lane together, warded the flank, or simply refused to approach from the front. Those are the behaviours that become more common as brackets rise, and they are the ones a hero that wants to stand still at maximum range has the least answer to.

The favourable list supports that. Lich, Zeus, Shadow Shaman and Queen of Pain are heroes that also want to hold a position and trade damage. Against opponents who agree to fight on those terms, Sniper’s range wins the argument comfortably. The decline by bracket is what happens when fewer opponents agree.

Two caveats. These matchups rest on 100 to 353 games each with four to eight points of uncertainty, so “nothing beats Sniper” means nothing beat it by enough to measure — a real counter with a three-point edge would be invisible at this sample size. And the bracket figures, built from tens of thousands of games apiece, are far firmer than any single pairing above them. Where the two disagree, trust the brackets.