Methodology

Where the data comes from

Blizzard publishes win, pick and ban rates for competitive play on its rates page. HeroStats reads the Americas region for PC and for Console, for every map and every rank, once a day, and keeps every daily snapshot. A hero's win rate is the share of their games won; pick rate is how often they are chosen; ban rate is how often they are banned before a match. Blizzard withholds a win rate when a sample is too small, and sometimes stops publishing per-map or ban data for a day or more; the site labels those cases rather than hiding them.

How tiers are scored

Each hero gets a score from three ingredients, then heroes are ranked within their role, because tanks and supports have intrinsically lower pick rates than damage heroes:

• Win rate above or below 50%, weighted ×4. This is the main signal.
• Pick rate, on a log scale, weighted ×8. A hero many players choose and still win with is stronger than a niche pick with the same win rate.
• Ban rate, square-rooted and weighted ×2, but only for heroes winning at or above 50%: full credit from 51%, none at 49%, a ramp between. Ban rate is corroborating evidence, not a standalone signal. Without that gate a hero at 48% win rate with a 40% ban rate scored as high as a genuinely strong hero on notoriety alone.

Scores are bucketed by how far they sit from the role's average: the average is the line between B and C, and each 0.75 standard deviations up or down is one tier (A/S above, D/F below). So S tier means "well clear of the pack in this role right now", not "always pick". Per-rank tier lists apply the same rule to that rank's numbers only.

Kit fit (✓✓ … ✗✗)

Every hero's abilities and perks are described in a small vocabulary — range bands, delivery (hitscan, projectile, beam), mobility kinds, barriers, self-sustain, knockback and so on — and every map by its mode and a few defining geometry traits (sniper lanes, long rotations, multi-level fights, enclosed interiors, environmental hazards). Rules connect kit traits to map traits: close-range brawl kits gain on compact Control points, long-range hitscan gains on sniper-lane Escorts, heroes with no way off the ground lose on multi-level maps, and so on. The glyph is the summed rule weight bucketed to five levels: ✓✓ great, ✓ good, – neutral, ✗ poor, ✗✗ bad.

Kit fit is independent of the observed numbers — it is what the kit implies, not what the win rate says — so a map where the two disagree is flagged rather than smoothed over. The rules are checked against the observed per-map win rates on both platforms across several days; they currently agree with the sign of the observed effect on about nine maps in ten, well above what knowing the map's mode alone predicts. Every rule and its measured effect is in the site's source.

Freshness

The footer shows two dates. Last scraped is when the site last read Blizzard's page. Numbers last changed is when those numbers last actually differed from the previous day — Blizzard refreshes the underlying data every two to three days, so identical days are expected. When Blizzard serves nothing for a rank on a given day, the previous day's numbers for that rank are shown with an "as of" note; when per-map or ban data is withheld, the last published values are shown with their date. A notice appears when the last scrape is more than four days old.

The roster evaluator

The roster page scores a pool of heroes against every map through both lenses: does someone in the pool suit the map (kit fit) and does someone in the pool win there (their win rate on that map versus their own average, at the selected rank). A map is covered when either lens says yes, a gap when neither does, and thin in between. Suggestions are the same-role heroes that would turn the most gaps into covered maps.