Methodology

Source and frequency

Every 10 minutes we check new replays for each active format and download them one by one, at most one per second. The log is stored as is and we extract what can be seen in the battle.

Sample bias

We only see battles someone saves publicly. It is a sample, not a census, and may lean towards players who upload their games. It is not the official Nintendo Switch game either.

What we count

The unit is the team: each side of a battle. In a best-of-three set, each player’s team counts once even if they play three games. Megas count as their base species and purely cosmetic forms (Vivillon, Alcremie...) are grouped.

Metrics

• Usage: share of teams that have the Pokémon in team preview.
• Brought: of the teams that have it, how many actually send it out (bring 6, pick 4).
• Lead: how many of those teams send it out on turn one.
• Moves: share of teams where the move is seen, over the teams where the Pokémon was sent out.
• Items and abilities: over the teams where one was revealed. We always show how many games it was revealed in.
• Teammates: share of its teams that also have the other Pokémon.
• Win rate: per game, with a 95% Wilson confidence interval.

Minimum sample

A Pokémon shows up in the ranking with at least 20 teams in the window. Move and item percentages need at least 30 appearances, and win rate needs at least 50 games. Below that, we show “not enough data”.

Windows

There are three windows: 24 hours, 7 days (default) and 30 days. The change (▲▼) compares with the previous window of the same length, in percentage points.

What we do not know

Replays do not show Stat Points or nature, so we do not publish full sets. Some battles have open team sheets; that data is stored apart and never mixed with what was revealed.