Overwatch tier list explained: how Bayesian shrinkage ranks every hero

Every row on the Counterwatch Overwatch tier list is sorted by shrunk win rate, not raw. That one distinction is what keeps the list useful when half the roster barely gets picked.

What the tiers measure

The tier list ranks every hero S through F by shrunk win rate, computed across community-tracked matches at your chosen mode, rank, and game type. S-tier means a hero is clearing the 50% baseline by a comfortable margin with enough matches behind the number to trust it. F-tier means the opposite, enough matches to be sure the hero is underperforming at that rank. The middle tiers, A through D, are where most of the roster sits. That is the nature of a game balanced around everyone being playable: the gap between the median damage hero and the S-tier one is not large at most ranks, and your own mastery usually matters more than a single tier of difference.

What shrinkage actually does

Here is the small-sample trap. A hero with five matches this week, all wins, has a 100% raw rate, and a naive sort drops them at the very top. That is obviously not real, and regression to the mean eats the number as soon as the sample grows. Shrinkage blends every hero's raw rate with a neutral 50% baseline, weighted by match count: thin samples get pulled most of the way to 50%, heavy samples barely budge. We add 400 imaginary coin-flip matches to each hero's record before computing the rate:

shrunk = (rawWinRate × matches + 0.5 × 400) / (matches + 400)

• 5 games at 100% raw becomes 50.6%. The fluke lands at baseline, nowhere near S-tier.
• 10,000 games at 55% raw becomes 54.8%. The real signal survives almost untouched.

A hero at 60% over 15 games reads the same as a 50.7% hero once shrinkage runs. To rank S-tier you need both a strong raw rate and enough data to back it.

Sample size and freshness

The shrunk rate on each row is computed over a rolling window of recent community matches. The site aggregates refresh once a day in the early UTC morning, so today's list reflects yesterday's games. We show every hero even when the sample is thin, but thin samples get shrunk hard, so a three-game fluke never floats to the top. Every hero page shows the match count behind the placement: tens of thousands is confident, a few hundred is provisional and can shift on any given day as the sample grows.

The rank filter is the lever most people skip

Bronze and Grandmaster are not the same game. Aim, game sense, and coordination all scale with rank, so heroes that demand precise mechanics or tight teamwork move tiers as you climb. The rank filter on the tier list runs from All through each division, Bronze, Silver, Gold, Platinum, Diamond, Masters, Grandmaster, Champion, plus a Grandmaster+ composite of the top two. Most people leave it on All and never notice how different their own rank looks. If you are in Silver, the Silver list is the one that matters for your climb; the All view smooths across every skill level and tells you what works on average, not what works in your games. The same page also filters by game type and by mode, so set all three to match what you actually queue.

Using it to climb

The tier list is a starting point, not gospel. Every row is an average across every map, every match, every lineup. It does not know which map you loaded, which comp the enemy locked, or which hero you have 800 hours on, and those decide your win rate on a given hero far more than one tier of placement does. So keep your mastery picks even when they sit a tier low, click through to a hero's page for the counters, synergies, and per-map splits where the real decisions live, and check best one tricks when you are choosing a main, since it weighs consistency and matchup breadth on top of raw win rate. In a live match the Counterwatch app overlay scores your options against the enemy lineup as picks lock, so you are not alt-tabbing to read a tier list from the spawn room. The methodology page has the full detail on shrinkage, the rolling window, and the per-rank data.