Brawl Atlas

Try out the draft agent and explore the visualizations based on millions of ranked games, updated weekly.

Rank all ranks

Rank range

Choose both ends of the range, or select a whole tier. Changes apply immediately.

Based on each lobby’s average rank, rounded down to its division. Data begins at Diamond III. Both endpoints are included.

Games in scope

Range

Who you're up against

Reading the lobby chart

Every ranked lobby in scope, bucketed by its average level. The highlighted bars are the slice your rank setting is reading from — move the slider and the rest of the page recounts from those lobbies.

Ranked tier is the in-game ladder: 22 steps from Bronze I through Masters III to Pro. Skill is the model's own estimate, as a z-score — 0 is an average lobby, ±1 is a standard deviation. The rank controls select the same lobbies on either chart. Skill is best compared across a full season, since ladder ranks shift as players climb.

Draft Agent

Backed by advanced machine learning and tree search strategies. Learn more

Your team

50%
even

Opponents

Bans

Bans

A ban takes a brawler off the board for both sides, the way ranked does above Mythic. Nobody can pick it and the search cannot answer with it.

This matters more than it looks. The agent assumes the opponent answers well, so the brawlers with the broadest answers shape every recommendation. Banning the few that are picked or banned in real games is usually the difference between a ranking that looks like the meta and one that looks like your lobby.

Draft strength

Draft strength

How many candidates each remaining pick weighs before choosing: Casual 4, Good 8, Sharp 16, Perfect every brawler left. The score is an exact expectation over that many draws rather than a sample, so it stays deterministic.

It applies to the whole lobby, teammates included. In solo queue you control one pick out of three, so assuming the rest of your team drafts as sharply as you do overstates your side. Raising it narrows the spread between candidates and pushes value toward brawlers that survive a precise answer; lowering it rewards brawlers that are simply strong on their own.

Best first picks

Search 1000 ms

Search time

How long the tree search runs. It expands the draft into a tree of replies, samples the lines that look strongest, and scores each candidate by the best line it finds beneath it.

At one ply no search runs: a pick is scored on the board it makes and nothing further. Longer budgets buy more simulations, which matter most on the opening picks, where five picks remain and the tree is spread thinnest. Estimates keep moving until roughly a second, then largely settle. The search runs on a worker thread, so the page stays responsive at any budget.

The Atlas

Reading the atlas

Color is strength: red is weaker and green is stronger. Size shows pick rate in the selected games. Drag to pan and use + / − to zoom. The numeric strength score is the model’s standalone score: higher is stronger, not a win probability.

Brawler positions come from learned embedding vectors built from all ranked games. Brawlers near each other play similarly; brawlers far apart play differently.

Positions stay fixed because the embedding vectors are based on all ranked games, not retrained for a map or Elo subset.

Grouped vs true distance

Grouped (t-SNE) — t-distributed stochastic neighbour embedding. It fits the 2-D picture so that each point keeps the same near neighbours it had in 64 dimensions, at the cost of the distances between clusters: it preserves local structure and deliberately discards the global scale. Good for "who plays like this brawler", misleading for "how far apart are these two groups".

True distance (PCA) — principal component analysis. It projects onto the two directions of greatest variance, a single linear map applied to everyone. Every distance you see is a real shadow of the distance in 64 dimensions, so the global geometry is honest — but the clusters overlap more, because two axes cannot hold what 64 were carrying.

100%

Strengthweakerstronger
Pick rateLowerHigher

Drag to pan · Positions stay fixed across filters

Brawler Data Table

⚠ Fewer than 100 games in this slice; rates may be noisy. Strength is a model score, not a win probability.

# Brawler Strength Win rate Picked Games