Marvel Rivals Methodology

TL;DR

• The tier list, counters, team-up synergies and ban rates are direct aggregates over recent 6v6 matches — what actually happened, not a model's opinion.
• The draft win predictor is live — pick both teams for a comp-based 6v6 win probability. It's calibrated (held-out ECE 0.014: a 60% call really wins ~60%), side-neutral (mirror comps read 50%), and honest about its ceiling: draft alone doesn't decide a hero-shooter, so most matchups sit near even.
• Everything is built on the current season only, so it reflects the live meta — not stale mixes of old patches. Every page shows its sample size, and rarer hero pairings still carry smaller samples, so you can weigh it yourself.

What's live

Where the data comes from

Public 6v6 match data is collected via the community marvelrivalsapi.com and rolled up into the snapshots these pages read. Each snapshot carries a generated_at timestamp and refreshes as new matches accumulate.

How each page is computed

• Tier list = each hero's win rate and pick rate over recent matches, also sliced into the four skill brackets above.
• Counters = a hero's win rate against each opposing hero (head-to-head, antisymmetric: if A beats B 56%, B beats A 44%).
• Team-up synergies = the win rate of two heroes on the same team (symmetric: the duo's joint win rate).
• Ban rates = how often each hero is removed in the draft phase, over matches that had a ban phase.
• Every hero and pairing shows its sample size, so a number backed by 4 games never looks like one backed by 400.

Sample sizes, in the open

Every number on these pages carries the sample behind it — here's the scale of the most recent snapshot. Rare hero pairings and the thinnest skill brackets still run on smaller samples; the per-row counts tell you which numbers to lean on.

Thresholds scale with the dataset; this box exists so you always know what the numbers stand on.

Known limitations

• Comp-only predictions. The draft predictor models the hero composition alone — player skill, map and execution are unmodeled, which is why most drafts honestly read near even.
• Thin tails. Win rates for rarely-played heroes, rare pairings and the thinnest skill brackets can swing as matches land; check the per-row sample sizes.
• Young game, shifting meta. Marvel Rivals patches and new heroes can move the meta faster than the data settles; fresh heroes sit near 50% until games accumulate.
• Not financial advice. This is an analytical / educational tool. The win rates here are not betting recommendations, and we do not endorse or facilitate sports betting.

Why bother, if it's just match data?

• Because it's honest match data, computed consistently — head-to-head and duo win rates, sliced by skill bracket, every number carrying its sample size.
• And because the draft predictor sits on top of this same data — the same pipeline behind Batru's other games, retrained weekly and calibrated on held-out current-season matches (ECE 0.014).