Overwatch ban recommendations: how to ban the right hero every match

A good Overwatch ban is not the best hero in the meta. It is the hero your specific lineup cannot beat, picked from the shortlist the enemy is most likely to run. Bans are also the one decision you cannot take back, which is what makes them harder than picks.

Why bans are harder than picks

Counter picks are forgiving. Lock the wrong one and you swap mid-match, costing thirty seconds while the enemy stacks the point. Bans happen once, pre-match, and they are final. Ban the wrong hero and the enemy runs exactly what you feared with an extra slot freed up. Ban the right one and they queue into a worse version of their comp while you keep full flexibility.

Most players ban on feel. "I hate fighting Widow, so I ban Widow." That is often wrong, because your team might have a solid Sombra and a good flex who both eat Widow alive, so banning her takes a hero off the table who was never going to threaten your specific lineup.

The two-factor ban score

Counterwatch's ban recommendation combines two things, how often a hero shows up and how badly they beat your team's picks when they do.

edge(opponent, friendly)  = max(0, rawWinRate(opp vs friendly) - 0.5)
edgeSum(opp)              = Σ edge across every friendly hero on your team
perMatchPickrate(opp)     = opponent's share of total tracked matches × heroes-per-match
threatScore(opp)          = perMatchPickrate(opp) × edgeSum(opp)

In plain terms: for each possible enemy hero, work out how much edge they have over each of your five, count only the matchups where they are winning, and sum those edges. A hero who badly beats two of your picks and is neutral into the other three outscores a hero who slightly beats all five. Then multiply by pickrate, because a hero who wrecks your team but almost never appears is still usually the wrong ban. The hero with the highest threat score is the recommendation, and both factors have to be high to top the list. A rare hero with a brutal edge and a common hero with a tiny edge both lose to a common hero with a real edge.

Why the ranking uses raw win rates

The tier list uses shrunk win rates because ranking is about fairness, and a hero with five tracked games at 100% does not deserve S-tier. A ban asks a different question, not whether a hero is fairly ranked but how much damage they do to your team if they show up. If a hero has a real 58% edge over your lineup across a few hundred games, the ban cares about that edge, not whether the sample is statistically perfect. Shrinking the edge would push the ranker toward banning popular-but-less-threatening heroes, which is the wrong call, so the threat score feeds on raw win rates from the counter table. Low-sample noise is handled separately. Any matchup with fewer than 200 tracked games is dropped before it reaches the ranking, rather than softened by shrinkage. Real 53% threats stay, fluky 65%-over-30-games threats never appear.

What the number on each row means

The win rate shown next to each ban candidate is the shrunk, match-weighted average, the same number you see for that opponent on the hardest-matchups card on the friendly hero's page. We show shrunk so you are not looking at two different numbers for the same matchup across the site, and above the 200-match floor raw and shrunk land within about a point of each other anyway, so the displayed number still reflects the threat the ranker responded to. Sorting follows the same logic. Switch the panel to Counter strength and it sorts by the displayed shrunk win rate so the column reads top to bottom, while the default Combined sort uses the raw-edge times pickrate threat score with the win-rate pill kept as context.

The shapes a good ban takes

Most right bans fall into one of a few shapes.

A high-pickrate baseline threat. Heroes like Ana, Kiriko, and Mercy show up as ban targets not because they crush matchups but because they are in nearly every game, so a small average edge times a 40 to 50% pickrate produces a steady threat score. These look boring and they are most often the top recommendation. If your team has no reliable answer to Ana, the recommendation often shows her first purely on pickrate, which is the formula doing its job.

A lineup-specific specialist. Less common and more dramatic, a hero with a 55%-plus edge into one of your picks, a Reaper into a Winston anchor or a Zarya into a Reinhardt, can rocket up the list even on a mid-tier pickrate, because the edge sum spikes when one or two friendly matchups are catastrophic. You will not see the same hero recommended for every lineup. If the tool surfaces a low-pickrate hero over a high-pickrate one, the edge-sum math has a reason.

A meta-enabler. Some heroes rank high because banning them denies a composition rather than a hero. Take Orisa off the table and you close out Orisa-anchored tank comps; ban Lúcio and you pull the speed boost out of many dive setups. The tool does not model "denies a comp" directly, but when the same high-pickrate hero keeps ranking high across several of your lineup options, there is usually a meta-enabler reason underneath.

Common ban mistakes

Banning on personal frustration is a tilt response, not a read, so check the recommendation. Banning what you expect instead of what you need wastes the ban when your team already counters the enemy's likely pick. Banning a hero only one teammate counters is backwards. If your whole team loses to Genji he is a great ban, but if one of your damage slots eats him alive, ban elsewhere. And do not ban a hero the enemy was unlikely to pick anyway, since low pickrate times any edge is close to noise.

Before you queue

The team builder runs the ban score live as you build a lineup, showing the top three candidates with the threat-score breakdown so you can see whether pickrate or edge is carrying each one, and the recommendations update the moment you change a friendly hero. For in-match bans where a mode allows them, the Counterwatch app overlay surfaces the same recommendations on live picks, so if the enemy locks a hero your team cannot beat and you still have a ban window, it names the hero. For how the score is built, the 200-match floor, the raw-versus-shrunk split, and why the ranking stays raw, see the methodology page.